From cc65abde0322acca870579dd3fdc56909620ae4a Mon Sep 17 00:00:00 2001
From: David Yastremsky <dyastremsky@nvidia.com>
Date: Mon, 26 Feb 2024 11:59:09 -0800
Subject: [PATCH 1/4] Copy over unittest files

---
 .../temp/qa/L0_unit_tests/count_tests.py      |    57 +
 .../temp/qa/L0_unit_tests/test.sh             |    50 +
 .../temp/qa/common/check_analyzer_results.sh  |   115 +
 .../temp/qa/common/check_copyright.py         |   239 +
 src/c++/perf_analyzer/temp/qa/common/util.sh  |   341 +
 src/c++/perf_analyzer/tests/__init__.py       |    15 +
 .../tests/common/bls-ckpt/0.ckpt              | 17358 ++++++++++++++++
 .../tests/common/bls-ckpt/README              |    15 +
 .../bls-ckpt/golden-metrics-model-gpu.csv     |    28 +
 .../golden-metrics-model-inference.csv        |    55 +
 .../bls-ckpt/golden-metrics-server-only.csv   |     3 +
 .../tests/common/ensemble-ckpt/0.ckpt         |  7257 +++++++
 .../tests/common/ensemble-ckpt/README         |    15 +
 .../golden-metrics-model-gpu.csv              |    23 +
 .../golden-metrics-model-inference.csv        |    23 +
 .../golden-metrics-server-only.csv            |     3 +
 .../tests/common/multi-model-ckpt/0.ckpt      |  7767 +++++++
 .../tests/common/multi-model-ckpt/README      |    19 +
 .../golden-metrics-model-gpu.csv              |    14 +
 .../golden-metrics-model-inference.csv        |    14 +
 .../golden-metrics-server-only.csv            |     3 +
 .../common/multi-model-ckpt/plot_manager.json |     1 +
 .../tests/common/request-rate-ckpt/0.ckpt     | 11428 ++++++++++
 .../tests/common/request-rate-ckpt/README     |    21 +
 .../golden-metrics-model-gpu.csv              |    36 +
 .../golden-metrics-model-inference.csv        |    36 +
 .../golden-metrics-server-only.csv            |     3 +
 .../tests/common/single-model-ckpt/0.ckpt     |  5350 +++++
 .../tests/common/single-model-ckpt/README     |    22 +
 .../golden-metrics-model-gpu.csv              |    18 +
 .../golden-metrics-model-inference.csv        |    18 +
 .../golden-metrics-server-only.csv            |     3 +
 .../single-model-ckpt/plot_manager.json       |     1 +
 .../tests/common/test_result_collector.py     |    47 +
 .../perf_analyzer/tests/common/test_utils.py  |   525 +
 src/c++/perf_analyzer/tests/mocks/__init__.py |    15 +
 .../tests/mocks/mock_api_error.py             |    24 +
 .../perf_analyzer/tests/mocks/mock_base.py    |    38 +
 .../perf_analyzer/tests/mocks/mock_client.py  |   221 +
 .../perf_analyzer/tests/mocks/mock_config.py  |    36 +
 .../perf_analyzer/tests/mocks/mock_dcgm.py    |    55 +
 .../tests/mocks/mock_dcgm_agent.py            |   197 +
 .../mocks/mock_dcgm_field_group_watcher.py    |    94 +
 .../perf_analyzer/tests/mocks/mock_glob.py    |    54 +
 .../tests/mocks/mock_gpu_device_factory.py    |    34 +
 src/c++/perf_analyzer/tests/mocks/mock_io.py  |   105 +
 .../perf_analyzer/tests/mocks/mock_json.py    |    61 +
 .../tests/mocks/mock_matplotlib.py            |    71 +
 .../tests/mocks/mock_model_config.py          |    49 +
 .../perf_analyzer/tests/mocks/mock_numba.py   |    66 +
 src/c++/perf_analyzer/tests/mocks/mock_os.py  |    91 +
 .../tests/mocks/mock_perf_analyzer.py         |   104 +
 .../perf_analyzer/tests/mocks/mock_psutil.py  |    35 +
 .../tests/mocks/mock_requests.py              |    54 +
 .../tests/mocks/mock_run_configs.py           |   144 +
 .../perf_analyzer/tests/mocks/mock_server.py  |    46 +
 .../tests/mocks/mock_server_docker.py         |   179 +
 .../tests/mocks/mock_server_local.py          |   131 +
 src/c++/perf_analyzer/tests/test_analyzer.py  |   241 +
 .../tests/test_bls_report_manager.py          |   172 +
 src/c++/perf_analyzer/tests/test_cli.py       |   604 +
 src/c++/perf_analyzer/tests/test_config.py    |  2382 +++
 .../tests/test_constraint_manager.py          |   620 +
 .../perf_analyzer/tests/test_coordinate.py    |   138 +
 .../tests/test_coordinate_data.py             |   123 +
 .../perf_analyzer/tests/test_cpu_monitor.py   |   111 +
 .../perf_analyzer/tests/test_dcgm_monitor.py  |   155 +
 .../tests/test_ensemble_report_manager.py     |   178 +
 .../perf_analyzer/tests/test_file_writer.py   |    67 +
 .../perf_analyzer/tests/test_html_report.py   |   167 +
 src/c++/perf_analyzer/tests/test_logger.py    |   175 +
 .../perf_analyzer/tests/test_model_config.py  |   472 +
 .../tests/test_model_config_generator.py      |   927 +
 .../tests/test_model_config_measurement.py    |   242 +
 .../tests/test_model_constraints.py           |   119 +
 .../perf_analyzer/tests/test_model_manager.py |  1319 ++
 .../tests/test_model_variant_name_manager.py  |   266 +
 .../perf_analyzer/tests/test_neighborhood.py  |   916 +
 .../tests/test_parameter_search.py            |   336 +
 .../perf_analyzer/tests/test_pdf_report.py    |    61 +
 .../perf_analyzer/tests/test_perf_analyzer.py |   873 +
 .../test_perf_analyzer_config_generator.py    |   842 +
 src/c++/perf_analyzer/tests/test_plot.py      |   131 +
 .../perf_analyzer/tests/test_plot_manager.py  |   196 +
 .../tests/test_quick_run_config_generator.py  |  1359 ++
 .../tests/test_record_aggregator.py           |   175 +
 .../perf_analyzer/tests/test_record_types.py  |   191 +
 .../tests/test_remote_monitor.py              |   174 +
 .../tests/test_report_manager.py              |   642 +
 .../perf_analyzer/tests/test_report_utils.py  |    60 +
 .../tests/test_result_manager.py              |   199 +
 .../perf_analyzer/tests/test_result_table.py  |   121 +
 .../tests/test_result_table_manager.py        |   344 +
 src/c++/perf_analyzer/tests/test_results.py   |   248 +
 .../perf_analyzer/tests/test_run_config.py    |   212 +
 .../tests/test_run_config_generator.py        |   729 +
 .../tests/test_run_config_measurement.py      |   739 +
 .../tests/test_run_config_result.py           |   341 +
 .../test_run_config_result_comparator.py      |   258 +
 .../perf_analyzer/tests/test_search_config.py |    88 +
 .../tests/test_search_dimension.py            |    53 +
 .../tests/test_search_dimensions.py           |    79 +
 .../tests/test_sorted_results.py              |   306 +
 .../perf_analyzer/tests/test_state_manager.py |   148 +
 .../perf_analyzer/tests/test_triton_client.py |   214 +
 .../perf_analyzer/tests/test_triton_server.py |   361 +
 .../tests/test_triton_server_factory.py       |   126 +
 .../perf_analyzer/tests/test_yaml_options.py  |   117 +
 108 files changed, 71944 insertions(+)
 create mode 100755 src/c++/perf_analyzer/temp/qa/L0_unit_tests/count_tests.py
 create mode 100755 src/c++/perf_analyzer/temp/qa/L0_unit_tests/test.sh
 create mode 100755 src/c++/perf_analyzer/temp/qa/common/check_analyzer_results.sh
 create mode 100755 src/c++/perf_analyzer/temp/qa/common/check_copyright.py
 create mode 100755 src/c++/perf_analyzer/temp/qa/common/util.sh
 create mode 100755 src/c++/perf_analyzer/tests/__init__.py
 create mode 100644 src/c++/perf_analyzer/tests/common/bls-ckpt/0.ckpt
 create mode 100644 src/c++/perf_analyzer/tests/common/bls-ckpt/README
 create mode 100644 src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-model-gpu.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-model-inference.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-server-only.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/ensemble-ckpt/0.ckpt
 create mode 100644 src/c++/perf_analyzer/tests/common/ensemble-ckpt/README
 create mode 100644 src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-model-gpu.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-model-inference.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-server-only.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/multi-model-ckpt/0.ckpt
 create mode 100644 src/c++/perf_analyzer/tests/common/multi-model-ckpt/README
 create mode 100644 src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-model-gpu.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-model-inference.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-server-only.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/multi-model-ckpt/plot_manager.json
 create mode 100644 src/c++/perf_analyzer/tests/common/request-rate-ckpt/0.ckpt
 create mode 100644 src/c++/perf_analyzer/tests/common/request-rate-ckpt/README
 create mode 100644 src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-model-gpu.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-model-inference.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-server-only.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/single-model-ckpt/0.ckpt
 create mode 100644 src/c++/perf_analyzer/tests/common/single-model-ckpt/README
 create mode 100644 src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-model-gpu.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-model-inference.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-server-only.csv
 create mode 100644 src/c++/perf_analyzer/tests/common/single-model-ckpt/plot_manager.json
 create mode 100755 src/c++/perf_analyzer/tests/common/test_result_collector.py
 create mode 100755 src/c++/perf_analyzer/tests/common/test_utils.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/__init__.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_api_error.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_base.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_client.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_config.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_dcgm.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_dcgm_agent.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_dcgm_field_group_watcher.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_glob.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_gpu_device_factory.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_io.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_json.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_matplotlib.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_model_config.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_numba.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_os.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_perf_analyzer.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_psutil.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_requests.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_run_configs.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_server.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_server_docker.py
 create mode 100755 src/c++/perf_analyzer/tests/mocks/mock_server_local.py
 create mode 100755 src/c++/perf_analyzer/tests/test_analyzer.py
 create mode 100755 src/c++/perf_analyzer/tests/test_bls_report_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_cli.py
 create mode 100755 src/c++/perf_analyzer/tests/test_config.py
 create mode 100755 src/c++/perf_analyzer/tests/test_constraint_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_coordinate.py
 create mode 100755 src/c++/perf_analyzer/tests/test_coordinate_data.py
 create mode 100755 src/c++/perf_analyzer/tests/test_cpu_monitor.py
 create mode 100755 src/c++/perf_analyzer/tests/test_dcgm_monitor.py
 create mode 100755 src/c++/perf_analyzer/tests/test_ensemble_report_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_file_writer.py
 create mode 100755 src/c++/perf_analyzer/tests/test_html_report.py
 create mode 100755 src/c++/perf_analyzer/tests/test_logger.py
 create mode 100755 src/c++/perf_analyzer/tests/test_model_config.py
 create mode 100755 src/c++/perf_analyzer/tests/test_model_config_generator.py
 create mode 100755 src/c++/perf_analyzer/tests/test_model_config_measurement.py
 create mode 100755 src/c++/perf_analyzer/tests/test_model_constraints.py
 create mode 100755 src/c++/perf_analyzer/tests/test_model_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_model_variant_name_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_neighborhood.py
 create mode 100755 src/c++/perf_analyzer/tests/test_parameter_search.py
 create mode 100755 src/c++/perf_analyzer/tests/test_pdf_report.py
 create mode 100755 src/c++/perf_analyzer/tests/test_perf_analyzer.py
 create mode 100755 src/c++/perf_analyzer/tests/test_perf_analyzer_config_generator.py
 create mode 100755 src/c++/perf_analyzer/tests/test_plot.py
 create mode 100755 src/c++/perf_analyzer/tests/test_plot_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_quick_run_config_generator.py
 create mode 100755 src/c++/perf_analyzer/tests/test_record_aggregator.py
 create mode 100755 src/c++/perf_analyzer/tests/test_record_types.py
 create mode 100755 src/c++/perf_analyzer/tests/test_remote_monitor.py
 create mode 100755 src/c++/perf_analyzer/tests/test_report_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_report_utils.py
 create mode 100755 src/c++/perf_analyzer/tests/test_result_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_result_table.py
 create mode 100755 src/c++/perf_analyzer/tests/test_result_table_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_results.py
 create mode 100755 src/c++/perf_analyzer/tests/test_run_config.py
 create mode 100755 src/c++/perf_analyzer/tests/test_run_config_generator.py
 create mode 100755 src/c++/perf_analyzer/tests/test_run_config_measurement.py
 create mode 100755 src/c++/perf_analyzer/tests/test_run_config_result.py
 create mode 100755 src/c++/perf_analyzer/tests/test_run_config_result_comparator.py
 create mode 100755 src/c++/perf_analyzer/tests/test_search_config.py
 create mode 100755 src/c++/perf_analyzer/tests/test_search_dimension.py
 create mode 100755 src/c++/perf_analyzer/tests/test_search_dimensions.py
 create mode 100755 src/c++/perf_analyzer/tests/test_sorted_results.py
 create mode 100755 src/c++/perf_analyzer/tests/test_state_manager.py
 create mode 100755 src/c++/perf_analyzer/tests/test_triton_client.py
 create mode 100755 src/c++/perf_analyzer/tests/test_triton_server.py
 create mode 100755 src/c++/perf_analyzer/tests/test_triton_server_factory.py
 create mode 100755 src/c++/perf_analyzer/tests/test_yaml_options.py

diff --git a/src/c++/perf_analyzer/temp/qa/L0_unit_tests/count_tests.py b/src/c++/perf_analyzer/temp/qa/L0_unit_tests/count_tests.py
new file mode 100755
index 000000000..d8bf8315d
--- /dev/null
+++ b/src/c++/perf_analyzer/temp/qa/L0_unit_tests/count_tests.py
@@ -0,0 +1,57 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+import importlib
+import inspect
+import os
+import sys
+
+sys.path.insert(0, "../../")
+
+
+def args():
+    parser = argparse.ArgumentParser("test_counter")
+    parser.add_argument("--path", help="Path to use for counting the tests", type=str)
+    opt = parser.parse_args()
+    return opt
+
+
+if __name__ == "__main__":
+    number_of_tests = 0
+    opt = args()
+    path = opt.path
+
+    for file_path in os.listdir(path):
+        # All the test files start with "Test"
+        if file_path.startswith("test_"):
+            module_name = "tests." + file_path.split(".")[0]
+            module = importlib.import_module(module_name)
+            classes = inspect.getmembers(module, inspect.isclass)
+            for class_tuple in classes:
+                class_name = class_tuple[0]
+                class_object = class_tuple[1]
+
+                # All the test classes start with "Test"
+                if class_name.startswith("Test"):
+                    methods = inspect.getmembers(class_object, inspect.isroutine)
+                    for method_tuple in methods:
+                        method_name = method_tuple[0]
+                        if method_name.startswith("test_"):
+                            number_of_tests += 1
+
+    # Print the number of tests
+    print(number_of_tests)
diff --git a/src/c++/perf_analyzer/temp/qa/L0_unit_tests/test.sh b/src/c++/perf_analyzer/temp/qa/L0_unit_tests/test.sh
new file mode 100755
index 000000000..b664ea1d4
--- /dev/null
+++ b/src/c++/perf_analyzer/temp/qa/L0_unit_tests/test.sh
@@ -0,0 +1,50 @@
+#!/bin/bash
+# Copyright (c) 2020-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+EXPECTED_NUM_TESTS=`python3 count_tests.py --path ../../tests/`
+source ../common/check_analyzer_results.sh
+source ../common/util.sh
+
+create_logs_dir "L0_unit_tests"
+create_result_paths -export-path false -checkpoints false
+
+RET=0
+
+set +e
+coverage run --branch --source=../../model_analyzer -m unittest discover -v -s ../../tests  -t ../../ > $ANALYZER_LOG 2>&1
+if [ $? -ne 0 ]; then
+    RET=1
+else
+    check_unit_test_results $ANALYZER_LOG $EXPECTED_NUM_TESTS
+    if [ $? -ne 0 ]; then
+        cat $ANALYZER_LOG
+        echo -e "\n***\n*** Test Result Verification Failed\n***"
+        RET=1
+    fi
+fi
+set -e
+
+if [ $RET -eq 0 ]; then
+    echo -e "\n***\n*** Test PASSED\n***"
+else
+    cat $ANALYZER_LOG
+    echo -e "\n***\n*** Test FAILED\n***"
+fi
+
+# Generate html files
+echo `pwd`
+coverage html --directory $LOGS_DIR/html/
+
+exit $RET
diff --git a/src/c++/perf_analyzer/temp/qa/common/check_analyzer_results.sh b/src/c++/perf_analyzer/temp/qa/common/check_analyzer_results.sh
new file mode 100755
index 000000000..79bc782d1
--- /dev/null
+++ b/src/c++/perf_analyzer/temp/qa/common/check_analyzer_results.sh
@@ -0,0 +1,115 @@
+#!/bin/bash
+# Copyright (c) 2021-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+source ../common/util.sh
+
+# Check Python unittest results.
+function check_unit_test_results () {
+    local log_file=$1
+    local expected_num_tests=$2
+
+    if [[ -z "$expected_num_tests" ]]; then
+        echo "=== expected number of tests must be defined"
+        return 1
+    fi
+
+    local num_failures=`cat $log_file | grep -E ".*total.*errors.*failures.*" | tail -n 1 | jq .failures`
+    local num_tests=`cat $log_file | grep -E ".*total.*errors.*failures.*" | tail -n 1 | jq .total`
+    local num_errors=`cat $log_file | grep -E ".*total.*errors.*failures.*" | tail -n 1 | jq .errors`
+
+    # Number regular expression
+    re='^[0-9]+$'
+
+    if [[ $? -ne 0 ]] || ! [[ $num_failures =~ $re ]] || ! [[ $num_tests =~ $re ]] || \
+     ! [[ $num_errors =~ $re ]]; then
+        cat $log_file
+        echo -e "\n***\n*** Test Failed: unable to parse test results\n***" >> $log_file
+        return 1
+    fi
+    if [[ $num_errors != "0" ]] || [[ $num_failures != "0" ]] || [[ $num_tests -ne $expected_num_tests ]]; then
+        cat $log_file
+        echo -e "\n***\n*** Test Failed: Expected $expected_num_tests test(s), $num_tests test(s) executed, $num_errors test(s) had error, and $num_failures test(s) failed. \n***" >> $log_file
+        return 1
+    fi
+
+    return 0
+}
+
+# Check the table(s) by calling check_log_table_row_column and check_csv_table_row_column
+# See if the logic generalizes to the specify testing needs
+# Call check_log_table_row_column and check_csv_table_row_column in util.sh directly if not
+function check_table_row_column() {
+    local inference_log_file=$1
+    local gpu_log_file=$2
+    local server_log_file=$3
+    local inference_csv_file=$4
+    local gpu_csv_file=$5
+    local server_csv_file=$6
+
+    local inference_expected_num_columns=$7
+    local inference_expected_num_rows=$8
+    local gpu_expected_num_columns=$9
+    local gpu_expected_num_rows=${10}
+    local server_expected_num_columns=${11}
+    local server_expected_num_rows=${12}
+
+    local ret=0
+
+    if [ "$inference_log_file" != "" ]; then
+        check_log_table_row_column $inference_log_file $inference_expected_num_columns $inference_expected_num_rows "Models\ \(Inference\):"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $inference_log_file.\n***"
+            ret=1
+        fi
+    fi
+    if [ "$gpu_log_file" != "" ]; then
+        check_log_table_row_column $gpu_log_file $gpu_expected_num_columns $gpu_expected_num_rows "Models\ \(GPU\ Metrics\):"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $gpu_log_file.\n***"
+            ret=1
+        fi
+    fi
+    if [ "$server_log_file" != "" ]; then
+        check_log_table_row_column $server_log_file $server_expected_num_columns $server_expected_num_rows "Server\ Only:"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $server_log_file.\n***"
+            ret=1
+        fi
+    fi
+
+    if [ "$inference_csv_file" != "" ]; then
+        check_csv_table_row_column $inference_csv_file $inference_expected_num_columns $inference_expected_num_rows "Model"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $inference_csv_file.\n***"
+            ret=1
+        fi
+    fi
+    if [ "$gpu_csv_file" != "" ]; then
+        check_csv_table_row_column $gpu_csv_file $gpu_expected_num_columns $gpu_expected_num_rows "Model"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $gpu_csv_file.\n***"
+            ret=1
+        fi
+    fi
+    if [ "$server_csv_file" != "" ]; then
+        check_csv_table_row_column $server_csv_file $server_expected_num_columns $server_expected_num_rows "Model"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $server_csv_file.\n***"
+            ret=1
+        fi
+    fi
+
+    return $ret
+}
diff --git a/src/c++/perf_analyzer/temp/qa/common/check_copyright.py b/src/c++/perf_analyzer/temp/qa/common/check_copyright.py
new file mode 100755
index 000000000..89c341512
--- /dev/null
+++ b/src/c++/perf_analyzer/temp/qa/common/check_copyright.py
@@ -0,0 +1,239 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import argparse
+import os
+import pathlib
+import re
+
+FLAGS = None
+SKIP_EXTS = ("pt", "log", "png", "pdf", "ckpt", "csv", "json")
+
+REPO_PATH_FROM_THIS_FILE = "../.."
+SKIP_PATHS = (".git", "VERSION", "LICENSE")
+
+COPYRIGHT_YEAR_RE = "Copyright( \\(c\\))? 20[1-9][0-9](-(20)?[1-9][0-9])?(,((20[2-9][0-9](-(20)?[2-9][0-9])?)|([2-9][0-9](-[2-9][0-9])?)))*,? NVIDIA CORPORATION( & AFFILIATES)?. All rights reserved."
+
+COPYRIGHT = """
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+"""
+
+repo_abs_path = (
+    pathlib.Path(__file__).parent.joinpath(REPO_PATH_FROM_THIS_FILE).resolve()
+)
+
+copyright_year_re = re.compile(COPYRIGHT_YEAR_RE)
+
+
+def visit(path):
+    if FLAGS.verbose:
+        print("visiting " + path)
+
+    for skip in SKIP_EXTS:
+        if path.endswith("." + skip):
+            if FLAGS.verbose:
+                print("skipping due to extension: " + path)
+            return True
+
+    for skip in SKIP_PATHS:
+        if str(pathlib.Path(path).resolve()).startswith(
+            str(repo_abs_path.joinpath(skip).resolve())
+        ):
+            if FLAGS.verbose:
+                print("skipping due to path prefix: " + path)
+            return True
+
+    with open(path, "r") as f:
+        first_line = True
+        line = None
+        try:
+            for fline in f:
+                line = fline
+
+                # Skip any '#!', '..', '<!--', or '{{/*' lines at the
+                # start of the file
+                if first_line:
+                    first_line = False
+                    if (
+                        fline.startswith("#!")
+                        or fline.startswith("..")
+                        or fline.startswith("<!--")
+                        or fline.startswith("{{/*")
+                    ):
+                        continue
+                # Skip empty lines...
+                if len(fline.strip()) != 0:
+                    break
+        except UnicodeDecodeError as ex:
+            # If we get this exception on the first line then assume a
+            # non-text file.
+            if not first_line:
+                raise ex
+            if FLAGS.verbose:
+                print("skipping binary file: " + path)
+            return True
+        if line is None:
+            if FLAGS.verbose:
+                print("skipping empty file: " + path)
+            return True
+
+        line = line.strip()
+
+        # The next line must be the copyright line with a single year
+        # or a year range. It is optionally allowed to have '# ' or
+        # '// ' prefix.
+        prefix = ""
+        if line.startswith("# "):
+            prefix = "# "
+        elif line.startswith("// "):
+            prefix = "// "
+        elif not line.startswith(COPYRIGHT_YEAR_RE[0]):
+            print(
+                "incorrect prefix for copyright line, allowed prefixes '# ' or '// ', for "
+                + path
+                + ": "
+                + line
+            )
+            return False
+
+        # Check if the copyright year line matches the regex
+        # and see if the year(s) are reasonable
+        years = []
+
+        copyright_row = line[len(prefix) :]
+        if copyright_year_re.match(copyright_row):
+            for year in (
+                copyright_row.split(
+                    "(c) " if "(c) " in copyright_row else "Copyright "
+                )[1]
+                .split(" NVIDIA ")[0]
+                .split(",")
+            ):
+                if len(year) == 4:  # 2021
+                    years.append(int(year))
+                elif len(year) == 2:  # 21
+                    years.append(int(year) + 2000)
+                elif len(year) == 9:  # 2021-2022
+                    years.append(int(year[0:4]))
+                    years.append(int(year[5:9]))
+                elif len(year) == 7:  # 2021-22
+                    years.append(int(year[0:4]))
+                    years.append(int(year[5:7]) + 2000)
+                elif len(year) == 5:  # 21-23
+                    years.append(int(year[0:2]) + 2000)
+                    years.append(int(year[3:5]) + 2000)
+        else:
+            print("copyright year is not recognized for " + path + ": " + line)
+            return False
+
+        if years[0] > FLAGS.year:
+            print(
+                "copyright start year greater than current year for "
+                + path
+                + ": "
+                + line
+            )
+            return False
+        if years[-1] > FLAGS.year:
+            print(
+                "copyright end year greater than current year for " + path + ": " + line
+            )
+            return False
+        for i in range(1, len(years)):
+            if years[i - 1] >= years[i]:
+                print("copyright years are not increasing for " + path + ": " + line)
+                return False
+
+        # Subsequent lines must match the copyright body.
+        copyright_body = [
+            l.rstrip() for i, l in enumerate(COPYRIGHT.splitlines()) if i > 0
+        ]
+        copyright_idx = 0
+        for line in f:
+            if copyright_idx >= len(copyright_body):
+                break
+
+            if len(prefix) == 0:
+                line = line.rstrip()
+            else:
+                line = line.strip()
+
+            if len(copyright_body[copyright_idx]) == 0:
+                expected = prefix.strip()
+            else:
+                expected = prefix + copyright_body[copyright_idx]
+            if line != expected:
+                print("incorrect copyright body for " + path)
+                print("  expected: '" + expected + "'")
+                print("       got: '" + line + "'")
+                return False
+            copyright_idx += 1
+
+        if copyright_idx != len(copyright_body):
+            print(
+                "missing "
+                + str(len(copyright_body) - copyright_idx)
+                + " lines of the copyright body"
+            )
+            return False
+
+    if FLAGS.verbose:
+        print("copyright correct for " + path)
+    return True
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "-v",
+        "--verbose",
+        action="store_true",
+        required=False,
+        default=False,
+        help="Enable verbose output",
+    )
+    parser.add_argument("-y", "--year", type=int, required=True, help="Copyright year")
+    parser.add_argument(
+        "paths", type=str, nargs="*", default=None, help="Directories or files to check"
+    )
+    FLAGS = parser.parse_args()
+
+    if FLAGS.paths is None or len(FLAGS.paths) == 0:
+        parser.print_help()
+        exit(1)
+
+    ret = True
+    for path in FLAGS.paths:
+        if not os.path.isdir(path):
+            if not visit(path):
+                ret = False
+        else:
+            for root, dirs, files in os.walk(path):
+                for name in files:
+                    if not visit(os.path.join(root, name)):
+                        ret = False
+
+    exit(0 if ret else 1)
diff --git a/src/c++/perf_analyzer/temp/qa/common/util.sh b/src/c++/perf_analyzer/temp/qa/common/util.sh
new file mode 100755
index 000000000..84054858a
--- /dev/null
+++ b/src/c++/perf_analyzer/temp/qa/common/util.sh
@@ -0,0 +1,341 @@
+#!/bin/bash
+# Copyright (c) 2020-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+LOGS_DIR=${LOGS_DIR:="/logs"}
+SERVER_LOG=${SERVER_LOG:="$LOGS_DIR/server.log"}
+SERVER_TIMEOUT=${SERVER_TIMEOUT:=120}
+SERVER_HTTP_PORT=${SERVER_HTTP_PORT:=8000}
+SERVER_LD_PRELOAD=${SERVER_LD_PRELOAD:=""}
+ANALYZER_LOG=${ANALYZER_LOG:="$LOGS_DIR/test.log"}
+
+mkdir -p $LOGS_DIR
+
+function create_logs_dir() {
+    # Arguments:
+    #  $1: L0 Script name
+    # Check if the L0 script name is empty or not
+    if [ -n "$1" ]; then
+        LOGS_DIR="/logs/$1"
+    else
+        LOGS_DIR="/logs"
+    fi
+    mkdir -p "$LOGS_DIR"
+}
+
+function create_result_paths() {
+    # Creates ANALYZER_LOG, EXPORT_PATH, CHECKPOINT_DIRECTORY and TEST_LOG_DIR
+    # Arguments:
+    #  -test-name <test_name>: <string> - L0 Script name (optional argument)
+    #  -export-path <false>: <boolean> - If false, skip creating EXPORT_PATH (optional argument)
+    #  -checkpoints <false>: <boolean> - If false, skip creating CHECKPOINT_DIRECTORY (optional argument)
+
+    # Set default values
+    test_name=""
+    export_path=true
+    checkpoints=true
+
+    # Parse arguments options
+    while [[ $# -gt 0 ]]; do
+        case $1 in
+            -test-name)
+                test_name=$2;
+                shift 2;;
+            -export-path)
+                export_path=$2;
+                shift 2;;
+            -checkpoints)
+                checkpoints=$2;
+                shift 2;;
+            *)
+                echo "Invalid option: $1" >&2;
+                return 1;;
+        esac
+    done
+
+    # Check if the test name is not an empty string
+    if [ -n "$test_name" ]; then
+        TEST_LOG_DIR="$LOGS_DIR/$test_name/logs"
+        ANALYZER_LOG="$TEST_LOG_DIR/analyzer.${test_name}.log"
+        mkdir -p "$TEST_LOG_DIR"
+
+        # Create EXPORT_PATH if export_path is true
+        if [ "$export_path" = true ]; then
+            EXPORT_PATH="$LOGS_DIR/$test_name/results"
+            mkdir -p "$EXPORT_PATH"
+        fi
+
+        # Create CHECKPOINT_DIRECTORY if checkpoints is true
+        if [ "$checkpoints" = true ]; then
+            CHECKPOINT_DIRECTORY="$LOGS_DIR/$test_name/checkpoints"
+            mkdir -p "$CHECKPOINT_DIRECTORY"
+        fi
+    else
+        TEST_LOG_DIR="$LOGS_DIR/logs"
+        ANALYZER_LOG="$TEST_LOG_DIR/analyzer.log"
+        mkdir -p "$TEST_LOG_DIR"
+
+        # Create EXPORT_PATH if export_path is true
+        if [ "$export_path" = true ]; then
+            EXPORT_PATH="$LOGS_DIR/results"
+            mkdir -p "$EXPORT_PATH"
+        fi
+
+        # Create CHECKPOINT_DIRECTORY if checkpoints is true
+        if [ "$checkpoints" = true ]; then
+            CHECKPOINT_DIRECTORY="$LOGS_DIR/checkpoints"
+            mkdir -p "$CHECKPOINT_DIRECTORY"
+        fi
+    fi
+}
+
+# Wait until server health endpoint shows ready. Sets WAIT_RET to 0 on
+# success, 1 on failure
+function wait_for_server_ready() {
+    local spid="$1"; shift
+    local wait_time_secs="${1:-30}"; shift
+
+    WAIT_RET=0
+
+    local wait_secs=$wait_time_secs
+    until test $wait_secs -eq 0 ; do
+        if ! kill -0 $spid; then
+            echo "=== Server not running."
+            WAIT_RET=1
+            return
+        fi
+
+        sleep 1;
+
+        set +e
+        code=`curl -s -w %{http_code} localhost:${1}/v2/health/ready`
+        set -e
+        if [ "$code" == "200" ]; then
+            return
+        fi
+
+        ((wait_secs--));
+    done
+
+    echo "=== Timeout $wait_time_secs secs. Server not ready."
+    WAIT_RET=1
+}
+
+# Run inference server. Return once server's health endpoint shows
+# ready or timeout expires. Sets SERVER_PID to pid of SERVER, or 0 if
+# error (including expired timeout)
+function run_server () {
+    SERVER_PID=0
+
+    if [ -z "$SERVER" ]; then
+        echo "=== SERVER must be defined"
+        return
+    fi
+
+    if [ ! -f "$SERVER" ]; then
+        echo "=== $SERVER does not exist"
+        return
+    fi
+
+    if [ -z "$SERVER_LD_PRELOAD" ]; then
+      echo "=== Running $SERVER $SERVER_ARGS"
+    else
+      echo "=== Running LD_PRELOAD=$SERVER_LD_PRELOAD $SERVER $SERVER_ARGS"
+    fi
+
+    LD_PRELOAD=$SERVER_LD_PRELOAD $SERVER $SERVER_ARGS > $SERVER_LOG 2>&1 &
+    SERVER_PID=$!
+
+    wait_for_server_ready $SERVER_PID $SERVER_TIMEOUT $SERVER_HTTP_PORT
+    if [ "$WAIT_RET" != "0" ]; then
+        kill $SERVER_PID || true
+        SERVER_PID=0
+    fi
+}
+
+# Run model-analyzer with args to completion.
+function run_analyzer() {
+    if [[ -z "$MODEL_ANALYZER" ]]; then
+        echo -e "=== model-analyzer executable not found"
+        return 1
+    fi
+
+    if [ ! -f "$MODEL_ANALYZER" ]; then
+        echo "=== $MODEL_ANALYZER does not exist"
+        return 1
+    fi
+
+    echo -e "=== Running $MODEL_ANALYZER $MODEL_ANALYZER_GLOBAL_OPTIONS $MODEL_ANALYZER_SUBCOMMAND $MODEL_ANALYZER_ARGS"
+    $MODEL_ANALYZER $MODEL_ANALYZER_GLOBAL_OPTIONS $MODEL_ANALYZER_SUBCOMMAND $MODEL_ANALYZER_ARGS >> $ANALYZER_LOG 2>&1
+    return $?
+}
+
+# Run model-analyzer with args to completion.
+function run_analyzer_nohup() {
+    if [[ -z "$MODEL_ANALYZER" ]]; then
+        echo -e "=== model-analyzer executable not found"
+        return 1
+    fi
+
+    if [ ! -f "$MODEL_ANALYZER" ]; then
+        echo "=== $MODEL_ANALYZER does not exist"
+        return 1
+    fi
+
+    echo -e "=== Running $MODEL_ANALYZER $MODEL_ANALYZER_GLOBAL_OPTIONS $MODEL_ANALYZER_SUBCOMMAND $MODEL_ANALYZER_ARGS"
+    $MODEL_ANALYZER $MODEL_ANALYZER_GLOBAL_OPTIONS $MODEL_ANALYZER_SUBCOMMAND $MODEL_ANALYZER_ARGS >> $ANALYZER_LOG 2>&1 &
+}
+
+# Check row and columns of csv file
+function check_csv_table_row_column() {
+    local csv_file=$1
+    local expected_num_columns=$2
+    local expected_num_rows=$3
+    local tag=$4
+    if [[ ! -f "$csv_file" ]]; then
+        echo -e "\n***\n*** Test Failed: $csv_file does not exist\n***"
+        return 1
+    fi
+
+    num_rows=`awk -v pattern=$tag '$0 ~ pattern {getline; i=0; while(getline) {i+=1}; print i}' $csv_file`
+    if [[ "$num_rows" != "$expected_num_rows" ]]; then
+        echo -e "\n***\n*** Test Failed: Expected $expected_num_rows rows in $csv_file, got ${num_rows}\n***"
+        echo -e "csv tag = ${tag}"
+        echo -e "csv csv_file = ${csv_file}"
+        echo -e "vvvvvvvvv"
+        cat ${csv_file}
+        echo -e "^^^^^^^^^"
+        return 1
+    fi
+    for i in $( seq 1 $expected_num_rows ); do
+        num_columns_found=`awk -v pattern=$tag -F ',' -v row="$i" '$0 ~ pattern {for (n=0; n<row;n++) {getline}; print NF}' $csv_file`
+        if [[ "$num_columns_found" != "$expected_num_columns" ]]; then
+            echo -e "\n***\n*** Test Failed: Expected $expected_num_columns columns in row $i, got ${num_columns_found}\n***"
+            echo -e "csv tag = ${tag}"
+            echo -e "csv row = ${i}"
+            echo -e "csv csv_file = ${csv_file}"
+            echo -e "vvvvvvvvv"
+            cat ${csv_file}
+            echo -e "^^^^^^^^^"
+            return 1
+        fi
+    done
+    return 0
+}
+
+function check_no_csv_exists() {
+    local csv_file=$1
+    if [[ ! -f "$csv_file" ]]; then
+        return 0
+    else
+        echo -e "\n***\n*** Test Failed: $csv_file should not exist\n***"
+        return 1
+    fi
+}
+
+function get_number_of_rows_logfile() {
+    local log_file=$1
+    num_rows_found=`awk  "BEGIN{i=0} /$tag/{flag=1;getline;getline} /^$/{flag=0} flag {i+=1} END{print i}" $log_file`
+    echo $number_row_found
+}
+
+function get_number_of_rows_csv() {
+    local csv_file=$1
+    num_rows=`awk -v pattern=$tag '$0 ~ pattern {getline; i=0; while(getline) {i+=1}; print i}' $csv_file`
+    echo $num_rows
+}
+
+# Check the output tables from the model-analyzer
+# This function simply ensures that there is a
+# Server-only table, a model table, and that
+# They have the specified dimensions
+function check_log_table_row_column() {
+    local log_file=$1
+    local expected_num_columns=$2
+    local expected_num_rows=$3
+    local tag=$4
+
+    # Check number of rows
+    num_rows_found=`awk  "BEGIN{i=0} /$tag/{flag=1;getline;getline} /^$/{flag=0} flag {i+=1} END{print i}" $log_file`
+    if [[ "$num_rows_found" != "$expected_num_rows" ]]; then
+        echo -e "\n***\n*** Test Failed: Expected $expected_num_rows rows $log_file, got ${num_rows_found}\n***"
+        echo -e "log tag = ${tag}"
+        echo -e "log log_file = ${log_file}"
+        echo -e "vvvvvvvvv"
+        cat ${log_file}
+        echo -e "^^^^^^^^^"
+        return 1
+    fi
+
+    # Check models table
+    for i in $( seq 1 $expected_num_rows ); do
+        # Columns in ith row
+        num_columns_found=`awk -v row="$i" "/$tag/{getline; for (n=0; n<row;n++) {getline}; print NF}" $log_file`
+        if [[ "$num_columns_found" != "$expected_num_columns" ]]; then
+            echo -e "\n***\n*** Test Failed: Expected $expected_num_columns columns in row $i of $tag, got ${num_columns_found}\n***"
+            echo -e "log tag = ${tag}"
+            echo -e "log row = ${i}"
+            echo -e "log log_file = ${log_file}"
+            echo -e "vvvvvvvvv"
+            cat ${log_file}
+            echo -e "^^^^^^^^^"
+            return 1
+        fi
+    done
+    return 0
+}
+
+function install_netstat() {
+    netstat > /dev/null 2>&1
+    if [ $? -eq 0 ]; then
+        return
+    else
+        DEBIAN_FRONTEND=noninteractive
+        apt-get update -qq > /dev/null 2>&1
+        DEBIAN_FRONTEND=noninteractive
+        apt-get install net-tools -y -qq > /dev/null 2>&1
+    fi
+}
+
+function find_available_ports() {
+    install_netstat
+    # First argument is the number of ports
+    num_port=$1
+
+    export ports=()
+    for i in `seq 1 $num_port`; do
+        current_port=$((10000 + $RANDOM % 10000))
+        while [ `netstat -ano tcp |& grep :$current_port > /dev/null 2>&1` ]; do
+            current_port=$(echo "$port + 1" | bc)
+        done
+        ports=("${ports[@]}" "$current_port")
+    done
+    echo ${ports[@]}
+}
+
+# Get a space delimited list of gpus with their UUIDs.
+function get_all_gpus_uuids() {
+    nvidia-smi --query-gpu=gpu_uuid,pci.bus_id --format=csv,noheader | awk 'BEGIN { FS = "," }{print $1}'
+}
+
+# We need to randomize the output directory so that in case the CI containers
+# end up on the same host, it does not create a conflict
+function get_output_directory() {
+    random_number=$((10000 + $RANDOM % 10000))
+    while [ `ls /tmp/output/ | grep $random_number` ]; do
+        random_number=$(echo "$random_number + 1" | bc)
+    done
+
+    echo /tmp/output/$random_number
+}
diff --git a/src/c++/perf_analyzer/tests/__init__.py b/src/c++/perf_analyzer/tests/__init__.py
new file mode 100755
index 000000000..f7f71eb3d
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/__init__.py
@@ -0,0 +1,15 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
diff --git a/src/c++/perf_analyzer/tests/common/bls-ckpt/0.ckpt b/src/c++/perf_analyzer/tests/common/bls-ckpt/0.ckpt
new file mode 100644
index 000000000..e43701de8
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/bls-ckpt/0.ckpt
@@ -0,0 +1,17358 @@
+{
+    "ResultManager.results": {
+        "_results": {
+            "bls": {
+                "bls_config_default": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "bls",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "bls",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "kind": "KIND_CPU"
+                                            }
+                                        ],
+                                        "backend": "python"
+                                    },
+                                    "variant_name": "bls_config_default",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": "add,sub"
+                                    },
+                                    "_options": {
+                                        "-m": "bls",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "bls-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "kind": "KIND_CPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_default",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "kind": "KIND_CPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_default",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "bls_config_default -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_default,sub_config_default": {
+                            "_model_variants_name": "bls_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.15,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.15,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.15,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.283,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.511,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.592,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.744,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 777.97,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.259,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.911,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.024,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.283,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.511,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.592,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.744,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 777.97,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.259,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.911,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.024,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_default -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_default,sub_config_default": {
+                            "_model_variants_name": "bls_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.025,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.025,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.025,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.398,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.643,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.769,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 832.413,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.372,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.99,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.024,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.102,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.398,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.643,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.769,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 832.413,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.372,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.99,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.024,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.102,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_default -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_default,sub_config_default": {
+                            "_model_variants_name": "bls_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.185,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.185,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.185,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 4.894,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.383,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.522,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 815.941,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 4.861,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 3.427,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.113,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 4.894,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.383,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.522,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 815.941,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 4.861,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 3.427,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.113,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_default -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_default,sub_config_default": {
+                            "_model_variants_name": "bls_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 34.2927,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 34.2927,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 34.2927,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 10.016,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 10.781,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 11.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 11.751,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 796.93,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 9.98,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 8.495,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.027,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.115,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 10.016,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 10.781,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 11.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 11.751,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 796.93,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 9.98,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 8.495,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.027,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.115,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_default -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_default,sub_config_default": {
+                            "_model_variants_name": "bls_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.2787,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.2787,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.2787,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 19.539,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 21.184,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 21.993,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 24.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 815.875,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 19.503,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 18.051,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.021,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.113,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 19.539,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 21.184,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 21.993,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 24.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 815.875,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 19.503,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 18.051,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.021,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.113,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "bls_config_0": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "bls",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "bls",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "backend": "python"
+                                    },
+                                    "variant_name": "bls_config_0",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": "add,sub"
+                                    },
+                                    "_options": {
+                                        "-m": "bls",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "bls-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_0",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_0",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "bls_config_0 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_0,sub_config_0": {
+                            "_model_variants_name": "bls_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.1707,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.1707,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.1707,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.352,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.676,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.77,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.06,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 849.002,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.327,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.984,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.983,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.025,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.107,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.352,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.676,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.77,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.06,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 849.002,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.327,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.984,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.983,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.025,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.107,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_0 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_0,sub_config_1": {
+                            "_model_variants_name": "bls_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.2007,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.2007,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.2007,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.364,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.662,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.759,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.995,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 844.557,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.337,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.981,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.981,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.111,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.364,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.662,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.759,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.995,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 844.557,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.337,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.981,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.981,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.111,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_0 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_2,sub_config_2": {
+                            "_model_variants_name": "bls_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6027,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6027,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6027,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.642,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.885,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.966,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.112,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 756.235,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.613,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.116,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.113,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.642,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.885,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.966,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.112,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 756.235,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.613,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.116,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.113,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "bls_config_1": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "bls",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "bls",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 2,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "backend": "python"
+                                    },
+                                    "variant_name": "bls_config_1",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": "add,sub"
+                                    },
+                                    "_options": {
+                                        "-m": "bls",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "bls-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 2,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_1",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_0",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "bls_config_1 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_1,sub_config_0": {
+                            "_model_variants_name": "bls_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.4793,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.4793,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.4793,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.15,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.229,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.413,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1316.47,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.241,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.275,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.15,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.229,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.413,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1316.47,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.241,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.275,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "bls_config_2": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "bls",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "bls",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 3,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "backend": "python"
+                                    },
+                                    "variant_name": "bls_config_2",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 6,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": "add,sub"
+                                    },
+                                    "_options": {
+                                        "-m": "bls",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "bls-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 2,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_1",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_0",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "bls_config_2 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=6 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_1,sub_config_0": {
+                            "_model_variants_name": "bls_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6313,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6313,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6313,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 6,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.199,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.492,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.591,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.884,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1874.23,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.164,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.32,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.346,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.199,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.492,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.591,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.884,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1874.23,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.164,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.32,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.346,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "bls_config_3": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "bls",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "bls",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 5,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "backend": "python"
+                                    },
+                                    "variant_name": "bls_config_3",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 10,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": 100,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": "add,sub"
+                                    },
+                                    "_options": {
+                                        "-m": "bls",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "bls-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 3,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_2",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_0",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "bls_config_3 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=10 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_2,sub_config_0": {
+                            "_model_variants_name": "bls_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6627,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6627,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6627,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 10,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.87,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 4.25,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 4.362,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.61,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2581.16,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.833,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.626,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.672,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.87,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 4.25,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 4.362,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.61,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2581.16,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.833,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.626,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.672,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_3 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=10 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_2,sub_config_1": {
+                            "_model_variants_name": "bls_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6107,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6107,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6107,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 10,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.853,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 4.221,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 4.334,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.591,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2591.54,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.816,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.622,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.664,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.853,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 4.221,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 4.334,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.591,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2591.54,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.816,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.622,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.664,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "bls_config_4": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "bls",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "bls",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 4,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "backend": "python"
+                                    },
+                                    "variant_name": "bls_config_4",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": "add,sub"
+                                    },
+                                    "_options": {
+                                        "-m": "bls",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "bls-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 4,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_3",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 3,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_2",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "bls_config_4 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_2": {
+                            "_model_variants_name": "bls_config_4",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.547,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.547,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.547,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.536,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.729,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.823,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2259.22,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.499,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.439,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.502,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.036,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.141,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.536,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.729,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.823,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2259.22,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.499,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 1.439,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.502,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.036,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.141,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "bls_config_5": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "bls",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "bls",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 6,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "backend": "python"
+                                    },
+                                    "variant_name": "bls_config_5",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 12,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": "add,sub"
+                                    },
+                                    "_options": {
+                                        "-m": "bls",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "bls-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 4,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_3",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_0",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=12 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.653,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.653,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.653,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 12,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.647,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.777,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.879,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.519,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 7284.42,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.633,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.708,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.726,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.647,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.777,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.879,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.519,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 7284.42,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.633,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.708,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.726,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=12 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_2": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.782,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.782,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.782,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 12,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.414,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.497,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.108,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 8475.91,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.401,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.573,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.607,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.052,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.414,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.497,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.108,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 8475.91,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.401,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.573,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.607,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.052,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=12 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_2": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.8133,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8133,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8133,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 12,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.362,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.479,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.531,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.826,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 8801.15,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.349,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.554,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.052,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.362,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.479,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.531,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.826,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 8801.15,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.349,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.554,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.052,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.4807,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.4807,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.4807,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.624,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.792,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.859,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.002,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 614.772,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.598,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.028,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.17,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.624,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.792,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.859,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.002,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 614.772,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.598,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.028,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.17,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.5833,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5833,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5833,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.793,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.934,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.018,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.27,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1113.32,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.762,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.027,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.276,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.793,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.934,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.018,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.27,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1113.32,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.762,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.027,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.276,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 34.6027,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 34.6027,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 34.6027,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.238,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.325,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.579,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1956.54,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.45,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.238,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.325,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.579,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1956.54,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.45,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.604,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.604,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.604,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.581,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.994,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.106,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.356,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3095.98,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.549,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.375,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.703,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.126,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.581,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.994,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.106,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.356,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3095.98,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.549,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.375,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.703,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.126,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.4258,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.4258,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.4258,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.306,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.994,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.429,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3014.95,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.272,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 3.037,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.747,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.13,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.306,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.994,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.429,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3014.95,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.272,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 3.037,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.747,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.13,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.2363,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.2363,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.2363,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 32,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 10.738,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 11.666,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 11.926,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 12.355,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2979.47,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 10.703,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 8.443,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.769,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.133,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 10.738,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 11.666,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 11.926,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 12.355,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2979.47,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 10.703,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 8.443,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.769,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.133,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=64 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.2308,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.2308,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.2308,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 64,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 19.18,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 22.205,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 22.844,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 23.781,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3328.92,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 19.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 17.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.581,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.028,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.118,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 19.18,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 22.205,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 22.844,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 23.781,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3328.92,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 19.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 17.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.581,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.028,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.118,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=128 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.1137,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.1137,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.1137,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 128,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 41.237,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 43.461,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 43.991,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 45.318,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3107.46,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 41.201,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 39.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.694,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.128,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 41.237,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 43.461,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 43.991,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 45.318,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3107.46,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 41.201,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 39.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.694,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.128,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=256 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.2747,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.2747,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.2747,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 256,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 60.083,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 82.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 83.071,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 86.796,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 4261.38,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 60.056,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 58.448,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.237,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.023,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.091,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 60.083,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 82.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 83.071,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 86.796,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 4261.38,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 60.056,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 58.448,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.237,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.023,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.091,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_5 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=512 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_0": {
+                            "_model_variants_name": "bls_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.225,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.225,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.225,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 512,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 108.534,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 141.676,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 152.157,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 165.976,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 4719.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 108.514,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 107.075,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.115,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.082,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 108.534,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 141.676,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 152.157,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 165.976,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 4719.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 108.514,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 107.075,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.115,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.082,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "bls_config_6": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "bls",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "bls",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 8,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "backend": "python"
+                                    },
+                                    "variant_name": "bls_config_6",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": "add,sub"
+                                    },
+                                    "_options": {
+                                        "-m": "bls",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "bls-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 6,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_4",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_0",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_0": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7087,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7087,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7087,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.216,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.836,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.372,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3066.01,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.183,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 2.325,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 2.377,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.125,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.216,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.836,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.372,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3066.01,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.183,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 2.325,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 2.377,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.125,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_1": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 53.4603,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 53.4603,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 53.4603,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.118,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.751,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.986,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.376,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3120.97,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.086,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 2.279,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 2.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.123,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.118,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.751,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.986,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.376,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3120.97,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.086,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 2.279,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 2.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.123,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_2": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.725,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.725,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.725,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.586,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.76,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.84,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.215,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 10077.3,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.573,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.624,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.702,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.586,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.76,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.84,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.215,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 10077.3,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.573,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.624,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.702,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_2,sub_config_4": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7035,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7035,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7035,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.627,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.761,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.848,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.367,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9824.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.613,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.642,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.721,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.048,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.627,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.761,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.848,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.367,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9824.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.613,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.642,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.721,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.048,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_0": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7623,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7623,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7623,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.719,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.916,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.011,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.177,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 580.858,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.691,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.028,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.253,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.142,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.719,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.916,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.011,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.177,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 580.858,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.691,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.028,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.253,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.142,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_0": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 34.7513,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 34.7513,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 34.7513,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.788,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.953,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.138,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1116.3,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.758,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.263,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.131,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.788,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.953,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.138,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1116.3,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.758,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.263,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.131,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_0": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.7323,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.7323,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.7323,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.069,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.26,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.508,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1929.86,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.426,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.128,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.069,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.26,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.508,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1929.86,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.426,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.128,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_0": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6932,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6932,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6932,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.876,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.987,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.392,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9115.99,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.863,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.876,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.987,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.392,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9115.99,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.863,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_0": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6878,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6878,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6878,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 32,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.223,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.423,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.595,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.679,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9915.61,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.209,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 2.249,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.711,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.051,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.223,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.423,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.595,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.679,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9915.61,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.209,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 2.249,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.711,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.051,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=64 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_0": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6623,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6623,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6623,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 64,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 6.5,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.76,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 7.109,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.155,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9822.01,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 6.486,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 5.511,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.717,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.052,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 6.5,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.76,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 7.109,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.155,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9822.01,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 6.486,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 5.511,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.717,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.052,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=128 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_0": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.639,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.639,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.639,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 128,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 13.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 13.185,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 14.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 18.842,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9772.16,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 13.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 11.922,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.719,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.015,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 13.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 13.185,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 14.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 18.842,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9772.16,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 13.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 11.922,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.719,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.015,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_6 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=256 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_0": {
+                            "_model_variants_name": "bls_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6405,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6405,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6405,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 256,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 26.088,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 26.526,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 28.456,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 33.974,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9731.24,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 26.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 25.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.723,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.015,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 26.088,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 26.526,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 28.456,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 33.974,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9731.24,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 26.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 25.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.723,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.015,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "bls_config_7": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "bls",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "bls",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 7,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "backend": "python"
+                                    },
+                                    "variant_name": "bls_config_7",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 14,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": "add,sub"
+                                    },
+                                    "_options": {
+                                        "-m": "bls",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "bls-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 5,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_6",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 3,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_2",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=14 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.632,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.632,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.632,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 14,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.526,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.632,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.717,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.236,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9161.43,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.513,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.623,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.661,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.055,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.526,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.632,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.717,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.236,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9161.43,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.513,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.623,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.661,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.055,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=14 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7988,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7988,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7988,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 14,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.455,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.593,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.645,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.776,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9610.87,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.442,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.591,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.628,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.455,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.593,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.645,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.776,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9610.87,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.442,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.591,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.628,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=14 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_4,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.8143,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8143,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8143,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 14,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.485,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.622,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.718,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.389,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9412.5,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.472,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.594,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.643,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.485,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.622,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.718,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.389,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9412.5,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.472,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.594,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.643,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=14 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_1": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.916,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.916,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.916,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 14,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.475,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.611,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.687,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.182,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9477.85,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.462,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.596,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.638,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.475,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.611,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.687,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.182,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9477.85,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.462,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.596,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.638,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=14 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_4": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.8205,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.8205,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.8205,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 14,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.467,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.597,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.657,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.893,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9532.59,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.453,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.593,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.632,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.467,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.597,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.657,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.893,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9532.59,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.453,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.593,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.632,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=14 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_2,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7725,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7725,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7725,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 14,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.452,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.587,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.636,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.752,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9629.36,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.439,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.578,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.635,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.048,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.452,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.587,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.636,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.752,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9629.36,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.439,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.578,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.635,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.048,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=14 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_1": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7557,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7557,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7557,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 14,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.473,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.612,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.678,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.021,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9489.7,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.46,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.585,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.643,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.473,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.612,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.678,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.021,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9489.7,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.46,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.585,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.643,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=14 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_3,sub_config_4": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 32.0528,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 32.0528,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 32.0528,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 14,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.464,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.597,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.656,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.914,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9551.42,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.45,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.639,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.464,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.597,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.656,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.914,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9551.42,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.45,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.639,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.735,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.735,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.735,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.699,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.913,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.981,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 587.452,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.672,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.237,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.142,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.699,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.913,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.981,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 587.452,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.672,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.237,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.142,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6767,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6767,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6767,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.794,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.968,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.018,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.153,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1113.08,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.763,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.027,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.262,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.127,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.794,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.968,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.018,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.153,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1113.08,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.008,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.763,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.027,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.262,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.127,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6833,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6833,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6833,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.066,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.253,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.325,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.464,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1932.8,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.426,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.036,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.129,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.066,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.253,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.325,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.464,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1932.8,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.009,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 1.426,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.036,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.129,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6703,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6703,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6703,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.877,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.997,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.051,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.254,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9099.48,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.864,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.609,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.877,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.997,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.051,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.254,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9099.48,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.864,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.609,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6685,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6685,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6685,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.689,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.874,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.938,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.167,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9460.29,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.676,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.811,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.638,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.055,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.689,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.874,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.938,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.167,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9460.29,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.676,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.811,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.638,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.055,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.7035,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.7035,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.7035,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 32,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.407,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.589,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.677,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.185,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9378.86,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.393,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 2.523,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.641,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.057,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 3.407,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.589,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.677,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 4.185,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9378.86,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 3.393,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 2.523,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.641,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.057,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "bls_config_7 -m bls -b 1 -i grpc -f bls-results.csv --verbose-csv --concurrency-range=64 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 --bls-composing-models=add,sub add_config_6,sub_config_2": {
+                            "_model_variants_name": "bls_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 23.6913,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6913,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 23.6913,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "bls",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 64,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 6.953,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 7.174,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 7.65,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.275,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9183.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 6.938,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 6.037,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.655,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.015,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.057,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 6.953,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 7.174,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 7.65,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.275,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 9183.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 6.938,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 6.037,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.655,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.015,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.057,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        }
+    },
+    "ResultManager.server_only_data": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+            [
+                "gpu_used_memory",
+                {
+                    "_value": 872.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_free_memory",
+                {
+                    "_value": 24897.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_utilization",
+                {
+                    "_value": 0.0,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ],
+            [
+                "gpu_power_usage",
+                {
+                    "_value": 56.0405,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ]
+        ]
+    },
+    "MetricsManager.gpus": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": {
+            "name": "NVIDIA TITAN RTX",
+            "total_memory": 25387401216
+        }
+    },
+    "ModelManager.model_variant_name_manager": {
+        "_model_config_dicts": {
+            "bls_config_0": {
+                "name": "bls",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "add_config_0": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "sub_config_0": {
+                "name": "sub",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "sub_config_1": {
+                "name": "sub",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "bls_config_1": {
+                "name": "bls",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "add_config_1": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "add_config_2": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "sub_config_2": {
+                "name": "sub",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "bls_config_2": {
+                "name": "bls",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "bls_config_3": {
+                "name": "bls",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 5,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "bls_config_4": {
+                "name": "bls",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 4,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "add_config_3": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 4,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "bls_config_5": {
+                "name": "bls",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 6,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "bls_config_6": {
+                "name": "bls",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 8,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "add_config_4": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 6,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "bls_config_7": {
+                "name": "bls",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 7,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "add_config_5": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 7,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "add_config_6": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 5,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "sub_config_3": {
+                "name": "sub",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 5,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "sub_config_4": {
+                "name": "sub",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 4,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            }
+        },
+        "_model_name_index": {
+            "bls": 7,
+            "add": 6,
+            "sub": 4
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/c++/perf_analyzer/tests/common/bls-ckpt/README b/src/c++/perf_analyzer/tests/common/bls-ckpt/README
new file mode 100644
index 000000000..b34f9cbd6
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/bls-ckpt/README
@@ -0,0 +1,15 @@
+# Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+This checkpoint was created by running the model from the BLS Quick Start.
diff --git a/src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-model-gpu.csv b/src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-model-gpu.csv
new file mode 100644
index 000000000..d72b48ce7
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-model-gpu.csv
@@ -0,0 +1,28 @@
+Model,GPU UUID,Batch,Concurrency,Model Config Path,Instance Group,Satisfies Constraints,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_7,8:GPU,Yes,1314.9,57.7,140.6
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,8,FaceDetectionBLS_config_7,8:GPU,Yes,1314.9,43.7,136.3
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,4,FaceDetectionBLS_config_7,8:GPU,Yes,1314.9,32.3,96.1
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,FaceDetectionBLS_config_7,8:GPU,Yes,1314.9,24.0,73.5
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,FaceDetectionBLS_config_7,8:GPU,Yes,1314.9,12.0,60.0
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_8,9:GPU,Yes,1314.9,28.0,114.0
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,8,FaceDetectionBLS_config_8,9:GPU,Yes,1314.9,54.3,118.5
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,4,FaceDetectionBLS_config_8,9:GPU,Yes,1314.9,52.7,118.2
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,FaceDetectionBLS_config_8,9:GPU,Yes,1314.9,23.0,77.1
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,FaceDetectionBLS_config_8,9:GPU,Yes,1314.9,9.7,60.1
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,8,FaceDetectionBLS_config_default,16:CPU,Yes,1312.8,42.0,142.4
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,4,FaceDetectionBLS_config_default,16:CPU,Yes,1312.8,30.7,100.0
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_default,16:CPU,Yes,1312.8,49.3,120.7
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,FaceDetectionBLS_config_default,16:CPU,Yes,1312.8,38.3,106.6
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,FaceDetectionBLS_config_default,16:CPU,Yes,1312.8,15.7,58.8
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_6,7:GPU,Yes,1314.9,40.7,111.5
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,4,FaceDetectionBLS_config_6,7:GPU,Yes,1314.9,31.7,98.8
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,8,FaceDetectionBLS_config_6,7:GPU,Yes,1314.9,50.7,120.2
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,FaceDetectionBLS_config_6,7:GPU,Yes,1314.9,23.3,81.5
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,FaceDetectionBLS_config_6,7:GPU,Yes,1314.9,10.7,60.5
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_5,6:GPU,Yes,1314.9,23.0,107.8
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_3,5:GPU,Yes,1314.9,17.3,103.0
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_4,4:GPU,Yes,1314.9,20.0,101.0
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_2,3:GPU,Yes,1312.8,32.3,95.7
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_1,2:GPU,Yes,1312.8,44.2,111.7
+FaceDetectionBLS,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,FaceDetectionBLS_config_0,1:GPU,Yes,1312.8,8.3,59.6
+
diff --git a/src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-model-inference.csv b/src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-model-inference.csv
new file mode 100644
index 000000000..fb69726d4
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-model-inference.csv
@@ -0,0 +1,55 @@
+Model,Batch,Concurrency,Model Config Path,Instance Group,Max Batch Size,Satisfies Constraints,Throughput (infer/sec),p99 Latency (ms)
+bls,1,16,bls_config_6,8:GPU,0,Yes,10077.3,2.2
+bls,1,32,bls_config_6,8:GPU,0,Yes,9915.6,4.7
+bls,1,16,bls_config_6,8:GPU,0,Yes,9824.1,2.4
+bls,1,64,bls_config_6,8:GPU,0,Yes,9822.0,10.2
+bls,1,128,bls_config_6,8:GPU,0,Yes,9772.2,18.8
+bls,1,256,bls_config_6,8:GPU,0,Yes,9731.2,34.0
+bls,1,8,bls_config_6,8:GPU,0,Yes,9116.0,1.4
+bls,1,16,bls_config_6,8:GPU,0,Yes,3121.0,6.4
+bls,1,16,bls_config_6,8:GPU,0,Yes,3066.0,6.4
+bls,1,4,bls_config_6,8:GPU,0,Yes,1929.9,2.5
+bls,1,2,bls_config_6,8:GPU,0,Yes,1116.3,2.1
+bls,1,1,bls_config_6,8:GPU,0,Yes,580.9,2.2
+bls,1,14,bls_config_7,7:GPU,0,Yes,9629.4,1.8
+bls,1,14,bls_config_7,7:GPU,0,Yes,9610.9,1.8
+bls,1,14,bls_config_7,7:GPU,0,Yes,9551.4,1.9
+bls,1,14,bls_config_7,7:GPU,0,Yes,9532.6,1.9
+bls,1,14,bls_config_7,7:GPU,0,Yes,9489.7,2.0
+bls,1,14,bls_config_7,7:GPU,0,Yes,9477.9,2.2
+bls,1,16,bls_config_7,7:GPU,0,Yes,9460.3,2.2
+bls,1,14,bls_config_7,7:GPU,0,Yes,9412.5,2.4
+bls,1,32,bls_config_7,7:GPU,0,Yes,9378.9,4.2
+bls,1,64,bls_config_7,7:GPU,0,Yes,9183.6,10.3
+bls,1,14,bls_config_7,7:GPU,0,Yes,9161.4,2.2
+bls,1,8,bls_config_7,7:GPU,0,Yes,9099.5,1.3
+bls,1,4,bls_config_7,7:GPU,0,Yes,1932.8,2.5
+bls,1,2,bls_config_7,7:GPU,0,Yes,1113.1,2.2
+bls,1,1,bls_config_7,7:GPU,0,Yes,587.5,2.1
+bls,1,12,bls_config_5,6:GPU,0,Yes,8801.1,1.8
+bls,1,12,bls_config_5,6:GPU,0,Yes,8475.9,2.1
+bls,1,12,bls_config_5,6:GPU,0,Yes,7284.4,2.5
+bls,1,512,bls_config_5,6:GPU,0,Yes,4719.6,166.0
+bls,1,256,bls_config_5,6:GPU,0,Yes,4261.4,86.8
+bls,1,64,bls_config_5,6:GPU,0,Yes,3328.9,23.8
+bls,1,128,bls_config_5,6:GPU,0,Yes,3107.5,45.3
+bls,1,8,bls_config_5,6:GPU,0,Yes,3096.0,3.4
+bls,1,16,bls_config_5,6:GPU,0,Yes,3014.9,6.4
+bls,1,32,bls_config_5,6:GPU,0,Yes,2979.5,12.4
+bls,1,4,bls_config_5,6:GPU,0,Yes,1956.5,2.6
+bls,1,2,bls_config_5,6:GPU,0,Yes,1113.3,2.3
+bls,1,1,bls_config_5,6:GPU,0,Yes,614.8,2.0
+bls,1,10,bls_config_3,5:GPU,0,Yes,2591.5,4.6
+bls,1,10,bls_config_3,5:GPU,0,Yes,2581.2,4.6
+bls,1,8,bls_config_4,4:GPU,0,Yes,2259.2,4.0
+bls,1,6,bls_config_2,3:GPU,0,Yes,1874.2,3.9
+bls,1,4,bls_config_1,2:GPU,0,Yes,1316.5,3.4
+bls,1,2,bls_config_0,1:GPU,0,Yes,849.0,3.1
+bls,1,2,bls_config_0,1:GPU,0,Yes,844.6,3.0
+bls,1,2,bls_config_0,1:GPU,0,Yes,756.2,3.1
+bls,1,2,bls_config_default,1:CPU,0,Yes,832.4,3.1
+bls,1,4,bls_config_default,1:CPU,0,Yes,815.9,6.0
+bls,1,16,bls_config_default,1:CPU,0,Yes,815.9,24.0
+bls,1,8,bls_config_default,1:CPU,0,Yes,796.9,11.8
+bls,1,1,bls_config_default,1:CPU,0,Yes,778.0,1.7
+
diff --git a/src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-server-only.csv b/src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-server-only.csv
new file mode 100644
index 000000000..77240adbf
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/bls-ckpt/golden-metrics-server-only.csv
@@ -0,0 +1,3 @@
+Model,GPU UUID,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+triton-server,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,872.0,0.0,56.0
+
diff --git a/src/c++/perf_analyzer/tests/common/ensemble-ckpt/0.ckpt b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/0.ckpt
new file mode 100644
index 000000000..2acf38be5
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/0.ckpt
@@ -0,0 +1,7257 @@
+{
+    "ResultManager.results": {
+        "_results": {
+            "ensemble_add_sub": {
+                "ensemble_add_sub_config_default": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "ensemble_add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "ensemble_add_sub",
+                                        "platform": "ensemble",
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "ensembleScheduling": {
+                                            "step": [
+                                                {
+                                                    "modelName": "add",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT1": "INPUT1",
+                                                        "INPUT0": "INPUT0"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT0": "OUTPUT0"
+                                                    }
+                                                },
+                                                {
+                                                    "modelName": "sub",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT1": "INPUT1",
+                                                        "INPUT0": "INPUT0"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT1": "OUTPUT1"
+                                                    }
+                                                }
+                                            ]
+                                        }
+                                    },
+                                    "variant_name": "ensemble_add_sub_config_default",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "ensemble_add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "ensemble_add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "kind": "KIND_CPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_default",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "kind": "KIND_CPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_default",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "ensemble_add_sub_config_default -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_default,sub_config_default": {
+                            "_model_variants_name": "ensemble_add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.321,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.321,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.321,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.565,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.639,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.697,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.866,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1764.33,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.544,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.324,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.126,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.565,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.639,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.697,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.866,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1764.33,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.544,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.324,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.126,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_default -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_default,sub_config_default": {
+                            "_model_variants_name": "ensemble_add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7245,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7245,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7245,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.706,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.817,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.864,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.949,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2825.37,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.683,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.333,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.706,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.817,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.864,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.949,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2825.37,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.683,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.333,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_default -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_default,sub_config_default": {
+                            "_model_variants_name": "ensemble_add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6613,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6613,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6613,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.4,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.616,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.695,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.879,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2851.58,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.374,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.346,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.045,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.4,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.616,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.695,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.879,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2851.58,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.374,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.346,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.045,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_default -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_default,sub_config_default": {
+                            "_model_variants_name": "ensemble_add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.648,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.648,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.648,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.89,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.362,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.499,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.77,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2763.39,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.86,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.365,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.89,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.362,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.499,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.77,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2763.39,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.86,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.365,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_default -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_default,sub_config_default": {
+                            "_model_variants_name": "ensemble_add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7327,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7327,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7327,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.731,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.499,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.683,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.159,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2786.78,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.7,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.361,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.731,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.499,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.683,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.159,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2786.78,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.7,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.361,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "ensemble_add_sub_config_0": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "ensemble_add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "ensemble_add_sub",
+                                        "platform": "ensemble",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "ensembleScheduling": {
+                                            "step": [
+                                                {
+                                                    "modelName": "add",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT1": "INPUT1",
+                                                        "INPUT0": "INPUT0"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT0": "OUTPUT0"
+                                                    }
+                                                },
+                                                {
+                                                    "modelName": "sub",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT1": "INPUT1",
+                                                        "INPUT0": "INPUT0"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT1": "OUTPUT1"
+                                                    }
+                                                }
+                                            ]
+                                        }
+                                    },
+                                    "variant_name": "ensemble_add_sub_config_0",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "ensemble_add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "ensemble_add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_0",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_0",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "ensemble_add_sub_config_0 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.5558,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5558,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5558,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.719,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.819,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.856,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.965,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2772.97,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.695,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.339,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.719,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.819,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.856,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.965,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2772.97,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.695,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.339,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_0 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7783,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7783,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7783,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.561,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.633,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.668,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.774,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1775.27,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.543,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.338,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.561,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.633,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.668,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.774,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1775.27,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.543,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.338,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_0 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6303,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6303,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6303,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.4,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.604,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.685,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.861,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2852.42,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.374,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.342,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.045,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.4,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.604,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.685,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.861,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2852.42,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.374,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.342,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.045,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_0 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7028,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7028,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7028,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.961,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.436,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.566,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.862,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2697.78,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.931,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.37,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.141,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.961,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.436,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.566,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.862,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2697.78,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.931,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.37,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.141,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_0 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 48.4545,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 48.4545,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 48.4545,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.758,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.525,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.729,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.174,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2773.69,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.728,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.362,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.139,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.758,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.525,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.729,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.174,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2773.69,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.728,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.362,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.139,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "ensemble_add_sub_config_1": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "ensemble_add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "ensemble_add_sub",
+                                        "platform": "ensemble",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "ensembleScheduling": {
+                                            "step": [
+                                                {
+                                                    "modelName": "add",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT0": "INPUT0",
+                                                        "INPUT1": "INPUT1"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT0": "OUTPUT0"
+                                                    }
+                                                },
+                                                {
+                                                    "modelName": "sub",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT0": "INPUT0",
+                                                        "INPUT1": "INPUT1"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT1": "OUTPUT1"
+                                                    }
+                                                }
+                                            ]
+                                        }
+                                    },
+                                    "variant_name": "ensemble_add_sub_config_1",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "ensemble_add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "ensemble_add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_0",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 2,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_1",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "ensemble_add_sub_config_1 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_1": {
+                            "_model_variants_name": "ensemble_add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.485,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.485,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.485,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.702,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.809,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.851,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.95,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2837.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.68,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.336,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.702,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.809,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.851,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.95,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2837.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.68,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.336,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_1 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_1": {
+                            "_model_variants_name": "ensemble_add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7663,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7663,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7663,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.558,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.647,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.693,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.794,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1785.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.541,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.131,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.558,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.647,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.693,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.794,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1785.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.541,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.131,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_1 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_1": {
+                            "_model_variants_name": "ensemble_add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.8155,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8155,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8155,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.559,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.808,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.887,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.064,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2561.85,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.528,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.382,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.559,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.808,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.887,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.064,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2561.85,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.528,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.382,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.049,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_1 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_1": {
+                            "_model_variants_name": "ensemble_add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6602,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6602,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6602,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.789,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.299,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.427,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.655,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2865.88,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.761,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.351,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.13,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.789,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.299,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.427,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.655,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2865.88,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.761,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.351,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.13,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_1 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_0,sub_config_1": {
+                            "_model_variants_name": "ensemble_add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 35.5865,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 35.5865,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 35.5865,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.701,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.596,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.819,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.253,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2801.72,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.671,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.359,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.701,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.596,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.819,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.253,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2801.72,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.671,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.359,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "ensemble_add_sub_config_2": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "ensemble_add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "ensemble_add_sub",
+                                        "platform": "ensemble",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "ensembleScheduling": {
+                                            "step": [
+                                                {
+                                                    "modelName": "add",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT1": "INPUT1",
+                                                        "INPUT0": "INPUT0"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT0": "OUTPUT0"
+                                                    }
+                                                },
+                                                {
+                                                    "modelName": "sub",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT1": "INPUT1",
+                                                        "INPUT0": "INPUT0"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT1": "OUTPUT1"
+                                                    }
+                                                }
+                                            ]
+                                        }
+                                    },
+                                    "variant_name": "ensemble_add_sub_config_2",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "ensemble_add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "ensemble_add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 2,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_1",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 1,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_0",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "ensemble_add_sub_config_2 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_1,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.5798,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5798,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5798,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.463,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.697,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.783,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.024,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2729.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.436,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.364,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.139,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.463,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.697,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.783,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.024,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2729.63,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.436,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.364,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.139,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_2 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_1,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.684,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.684,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.684,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.563,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.637,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.678,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.777,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1770.93,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.543,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.563,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.637,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.678,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.777,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1770.93,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.543,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_2 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_1,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7113,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7113,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7113,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.7,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.802,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.846,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.938,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2847.24,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.677,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.334,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.7,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.802,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.846,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.938,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2847.24,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.677,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.334,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_2 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_1,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.641,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.641,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.641,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.888,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.357,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.497,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.749,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2766.53,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.858,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.364,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.888,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 3.357,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 3.497,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 3.749,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2766.53,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.858,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.364,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "ensemble_add_sub_config_2 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_1,sub_config_0": {
+                            "_model_variants_name": "ensemble_add_sub_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.795,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.795,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.795,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 16,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.91,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.698,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.906,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.706,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2702.59,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.88,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.371,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.91,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.698,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.906,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.706,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2702.59,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.007,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.88,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.371,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "ensemble_add_sub_config_3": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "ensemble_add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "ensemble_add_sub",
+                                        "platform": "ensemble",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            },
+                                            {
+                                                "name": "OUTPUT1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "4"
+                                                ]
+                                            }
+                                        ],
+                                        "ensembleScheduling": {
+                                            "step": [
+                                                {
+                                                    "modelName": "add",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT1": "INPUT1",
+                                                        "INPUT0": "INPUT0"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT0": "OUTPUT0"
+                                                    }
+                                                },
+                                                {
+                                                    "modelName": "sub",
+                                                    "modelVersion": "1",
+                                                    "inputMap": {
+                                                        "INPUT0": "INPUT0",
+                                                        "INPUT1": "INPUT1"
+                                                    },
+                                                    "outputMap": {
+                                                        "OUTPUT1": "OUTPUT1"
+                                                    }
+                                                }
+                                            ]
+                                        }
+                                    },
+                                    "variant_name": "ensemble_add_sub_config_3",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 6,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "ensemble_add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "ensemble_add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": [
+                                    {
+                                        "model_config": {
+                                            "name": "add",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 3,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "add_config_2",
+                                        "cpu_only": false
+                                    },
+                                    {
+                                        "model_config": {
+                                            "name": "sub",
+                                            "input": [
+                                                {
+                                                    "name": "INPUT0",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                },
+                                                {
+                                                    "name": "INPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "output": [
+                                                {
+                                                    "name": "OUTPUT1",
+                                                    "dataType": "TYPE_FP32",
+                                                    "dims": [
+                                                        "4"
+                                                    ]
+                                                }
+                                            ],
+                                            "instanceGroup": [
+                                                {
+                                                    "count": 3,
+                                                    "kind": "KIND_GPU"
+                                                }
+                                            ],
+                                            "backend": "python"
+                                        },
+                                        "variant_name": "sub_config_2",
+                                        "cpu_only": false
+                                    }
+                                ]
+                            }
+                        ]
+                    },
+                    {
+                        "ensemble_add_sub_config_3 -m ensemble_add_sub -b 1 -i grpc -f ensemble_add_sub-results.csv --verbose-csv --concurrency-range=6 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000 add_config_2,sub_config_2": {
+                            "_model_variants_name": "ensemble_add_sub_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 870.31808,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24899.485696,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6983,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6983,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 870.31808,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24899.485696,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6983,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "ensemble_add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 6,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.211,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.534,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.635,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.835,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2709.16,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.183,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.371,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 2.211,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 2.534,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 2.635,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 2.835,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2709.16,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.006,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 2.183,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.001,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.371,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        }
+    },
+    "ResultManager.server_only_data": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+            [
+                "gpu_used_memory",
+                {
+                    "_value": 870.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_free_memory",
+                {
+                    "_value": 24899.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_utilization",
+                {
+                    "_value": 0.0,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ],
+            [
+                "gpu_power_usage",
+                {
+                    "_value": 56.084,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ]
+        ]
+    },
+    "MetricsManager.gpus": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": {
+            "name": "NVIDIA TITAN RTX",
+            "total_memory": 25387401216
+        }
+    },
+    "ModelManager.model_variant_name_manager": {
+        "_model_config_dicts": {
+            "add_config_0": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "sub_config_0": {
+                "name": "sub",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "ensemble_add_sub_config_0": {
+                "key": "add_config_0,sub_config_0",
+                "name": "ensemble_add_sub"
+            },
+            "sub_config_1": {
+                "name": "sub",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "ensemble_add_sub_config_1": {
+                "key": "add_config_0,sub_config_1",
+                "name": "ensemble_add_sub"
+            },
+            "add_config_1": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "ensemble_add_sub_config_2": {
+                "key": "add_config_1,sub_config_0",
+                "name": "ensemble_add_sub"
+            },
+            "add_config_2": {
+                "name": "add",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "sub_config_2": {
+                "name": "sub",
+                "input": [
+                    {
+                        "name": "INPUT0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    },
+                    {
+                        "name": "INPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "4"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "backend": "python"
+            },
+            "ensemble_add_sub_config_3": {
+                "key": "add_config_2,sub_config_2",
+                "name": "ensemble_add_sub"
+            }
+        },
+        "_model_name_index": {
+            "add": 2,
+            "sub": 2,
+            "ensemble_add_sub": 3
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/c++/perf_analyzer/tests/common/ensemble-ckpt/README b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/README
new file mode 100644
index 000000000..8e8fa2b11
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/README
@@ -0,0 +1,15 @@
+# Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+This checkpoint was created by running the Ensemble Quick Start.
diff --git a/src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-model-gpu.csv b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-model-gpu.csv
new file mode 100644
index 000000000..b6af22d8c
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-model-gpu.csv
@@ -0,0 +1,23 @@
+Model,GPU UUID,Batch,Concurrency,Model Config Path,Instance Group,Satisfies Constraints,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,8,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU",Yes,870.3,0.0,56.7
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU",Yes,870.3,0.0,56.5
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU",Yes,870.3,0.0,35.6
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,4,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU",Yes,870.3,0.0,56.8
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU",Yes,870.3,0.0,56.8
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,4,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU",Yes,870.3,0.0,56.6
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU",Yes,870.3,0.0,48.5
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU",Yes,870.3,0.0,56.6
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,8,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU",Yes,870.3,0.0,56.7
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU",Yes,870.3,0.0,56.8
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,4,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU",Yes,870.3,0.0,56.7
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU",Yes,870.3,0.0,56.7
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU",Yes,870.3,0.0,56.7
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,8,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU",Yes,870.3,0.0,56.6
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU",Yes,870.3,0.0,56.3
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU",Yes,870.3,0.0,56.7
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,8,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU",Yes,870.3,0.0,56.6
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,4,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU",Yes,870.3,0.0,56.6
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU",Yes,870.3,0.0,56.8
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU",Yes,870.3,0.0,56.7
+ensemble_add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,6,"ensemble_add_sub_config_3: add_config_2, sub_config_2","3:GPU,3:GPU",Yes,870.3,0.0,56.7
+
diff --git a/src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-model-inference.csv b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-model-inference.csv
new file mode 100644
index 000000000..18444821f
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-model-inference.csv
@@ -0,0 +1,23 @@
+Model,Batch,Concurrency,Model Config Path,Instance Group,Max Batch Size,Satisfies Constraints,Throughput (infer/sec),p99 Latency (ms)
+ensemble_add_sub,1,8,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU","0,0",Yes,2865.9,3.7
+ensemble_add_sub,1,2,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU","0,0",Yes,2837.6,0.9
+ensemble_add_sub,1,16,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU","0,0",Yes,2801.7,7.3
+ensemble_add_sub,1,4,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU","0,0",Yes,2561.8,2.1
+ensemble_add_sub,1,1,"ensemble_add_sub_config_1: add_config_0, sub_config_1","1:GPU,2:GPU","0,0",Yes,1785.0,0.8
+ensemble_add_sub,1,4,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU","0,0",Yes,2852.4,1.9
+ensemble_add_sub,1,16,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU","0,0",Yes,2773.7,7.2
+ensemble_add_sub,1,2,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU","0,0",Yes,2773.0,1.0
+ensemble_add_sub,1,8,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU","0,0",Yes,2697.8,3.9
+ensemble_add_sub,1,1,"ensemble_add_sub_config_0: add_config_0, sub_config_0","1:GPU,1:GPU","0,0",Yes,1775.3,0.8
+ensemble_add_sub,1,4,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU","0,0",Yes,2851.6,1.9
+ensemble_add_sub,1,2,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU","0,0",Yes,2825.4,0.9
+ensemble_add_sub,1,16,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU","0,0",Yes,2786.8,7.2
+ensemble_add_sub,1,8,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU","0,0",Yes,2763.4,3.8
+ensemble_add_sub,1,1,"ensemble_add_sub_config_default: add_config_default, sub_config_default","1:CPU,1:CPU","0,0",Yes,1764.3,0.9
+ensemble_add_sub,1,2,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU","0,0",Yes,2847.2,0.9
+ensemble_add_sub,1,8,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU","0,0",Yes,2766.5,3.7
+ensemble_add_sub,1,4,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU","0,0",Yes,2729.6,2.0
+ensemble_add_sub,1,16,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU","0,0",Yes,2702.6,7.7
+ensemble_add_sub,1,1,"ensemble_add_sub_config_2: add_config_1, sub_config_0","2:GPU,1:GPU","0,0",Yes,1770.9,0.8
+ensemble_add_sub,1,6,"ensemble_add_sub_config_3: add_config_2, sub_config_2","3:GPU,3:GPU","0,0",Yes,2709.2,2.8
+
diff --git a/src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-server-only.csv b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-server-only.csv
new file mode 100644
index 000000000..4daa8fc3d
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/ensemble-ckpt/golden-metrics-server-only.csv
@@ -0,0 +1,3 @@
+Model,GPU UUID,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+triton-server,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,870.0,0.0,56.1
+
diff --git a/src/c++/perf_analyzer/tests/common/multi-model-ckpt/0.ckpt b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/0.ckpt
new file mode 100644
index 000000000..a515cc495
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/0.ckpt
@@ -0,0 +1,7767 @@
+{
+    "ResultManager.results": {
+        "_results": {
+            "resnet50_libtorch,vgg19_libtorch": {
+                "resnet50_libtorch_config_default,vgg19_libtorch_config_default": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "resnet50_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "resnet50_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 128,
+                                        "input": [
+                                            {
+                                                "name": "input",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "output",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "resnet50_labels.txt"
+                                            }
+                                        ]
+                                    },
+                                    "variant_name": "resnet50_libtorch_config_default",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "resnet50_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "resnet50_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            },
+                            {
+                                "_model_name": "vgg19_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "vgg19_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 128,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "vgg19_labels.txt"
+                                            }
+                                        ]
+                                    },
+                                    "variant_name": "vgg19_libtorch_config_default",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "vgg19_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "vgg19_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "resnet50_libtorch_config_default -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_default -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_default,vgg19_libtorch_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2296.38144,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23473.422336,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 99.66669999999999,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 273.771,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 99.66669999999999,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 273.771,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 99.66669999999999,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 273.771,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.956,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.607,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.848,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.683,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 167.896,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.84,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.016,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.986,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.123,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.956,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.607,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.848,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.683,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 167.896,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.84,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.016,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.986,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.123,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.665,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.79,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.813,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 5.866,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 176.223,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.552,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.016,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.742,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.122,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.665,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.79,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.813,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 5.866,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 176.223,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.552,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.016,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.742,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.122,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "resnet50_libtorch_config_default -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_default -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_default,vgg19_libtorch_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2296.38144,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23473.422336,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 99.5,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 277.904,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 99.5,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 277.904,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 99.5,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 277.904,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 14.46,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 15.086,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 15.251,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 15.622,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 138.552,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 14.333,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 6.453,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.99,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.146,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 14.46,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 15.086,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 15.251,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 15.622,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 138.552,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 14.333,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 6.453,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.99,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.146,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 9.971,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 10.166,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 10.223,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.341,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 200.5,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 9.855,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 4.266,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.765,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 9.971,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 10.166,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 10.223,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.341,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 200.5,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 9.855,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 4.266,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.765,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "resnet50_libtorch_config_default -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_default -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_default,vgg19_libtorch_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2296.38144,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23473.422336,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 50.5,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 178.015,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 50.5,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 178.015,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 50.5,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 178.015,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 28.594,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 29.911,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 30.124,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 30.448,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 139.233,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.11,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 28.456,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 20.613,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.926,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 28.594,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 29.911,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 30.124,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 30.448,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 139.233,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.11,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 28.456,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 20.613,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.926,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 19.973,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 20.289,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 20.419,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 20.683,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 199.529,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.105,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 19.841,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 14.17,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.767,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.145,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 19.973,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 20.289,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 20.419,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 20.683,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 199.529,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.105,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 19.841,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 14.17,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.767,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.145,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "resnet50_libtorch_config_default -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_default -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_default,vgg19_libtorch_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2296.38144,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23473.422336,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 60.3333,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 216.06,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 60.3333,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 216.06,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 60.3333,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 216.06,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 57.152,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 59.898,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 60.204,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 60.672,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 138.549,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.119,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 57.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 49.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.953,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 57.152,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 59.898,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 60.204,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 60.672,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 138.549,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.119,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 57.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 49.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.953,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 39.855,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 40.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 40.389,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 40.655,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 199.507,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 39.68,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 33.962,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.769,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 39.855,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 40.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 40.389,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 40.655,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 199.507,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 39.68,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 33.962,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.769,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "resnet50_libtorch_config_0,vgg19_libtorch_config_0": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "resnet50_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "resnet50_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "input",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "output",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "resnet50_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "resnet50_libtorch_config_0",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "resnet50_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "resnet50_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            },
+                            {
+                                "_model_name": "vgg19_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "vgg19_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "vgg19_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "vgg19_libtorch_config_0",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "vgg19_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "vgg19_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "resnet50_libtorch_config_0 -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_0 -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_0,vgg19_libtorch_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2296.38144,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23473.422336,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 97.66669999999999,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 269.32,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 97.66669999999999,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 269.32,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2296.38144,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23473.422336,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 97.66669999999999,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 269.32,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 14.31,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 15.01,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 15.154,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 15.455,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 139.554,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.105,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 14.177,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 6.358,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.915,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.144,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 14.31,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 15.01,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 15.154,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 15.455,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 139.554,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.105,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 14.177,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 6.358,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.915,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.144,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 9.907,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 10.113,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 10.18,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.32,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 201.837,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 9.78,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 4.181,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.729,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 9.907,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 10.113,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 10.18,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.32,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 201.837,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 9.78,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 4.181,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.729,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "resnet50_libtorch_config_0,vgg19_libtorch_config_4": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "resnet50_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "resnet50_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "input",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "output",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "resnet50_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "resnet50_libtorch_config_0",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "resnet50_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "resnet50_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            },
+                            {
+                                "_model_name": "vgg19_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "vgg19_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "vgg19_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "vgg19_libtorch_config_4",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "vgg19_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "vgg19_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "resnet50_libtorch_config_0 -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_4 -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_0,vgg19_libtorch_config_4",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2323.644416,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23446.15936,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 99.3333,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 275.388,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2323.644416,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23446.15936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 99.3333,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 275.388,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2323.644416,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23446.15936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 99.3333,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 275.388,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 18.466,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 19.367,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 19.625,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 19.916,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 108.59,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.107,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 18.331,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 8.443,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 8.977,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 18.466,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 19.367,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 19.625,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 19.916,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 108.59,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.107,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 18.331,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 8.443,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 8.977,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 15.394,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 15.651,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 15.727,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 15.803,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 259.817,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.124,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 15.243,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 6.726,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 7.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.28,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 15.394,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 15.651,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 15.727,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 15.803,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 259.817,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.124,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 15.243,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 6.726,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 7.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.28,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "resnet50_libtorch_config_7,vgg19_libtorch_config_1": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "resnet50_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "resnet50_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "input",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "output",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "resnet50_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "resnet50_libtorch_config_7",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "resnet50_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "resnet50_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            },
+                            {
+                                "_model_name": "vgg19_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "vgg19_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "vgg19_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 2,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "vgg19_libtorch_config_1",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "vgg19_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "vgg19_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "resnet50_libtorch_config_7 -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_1 -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_7,vgg19_libtorch_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 3089.104896,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 22680.69888,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 79.6667,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 280.57,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 3089.104896,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 22680.69888,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 79.6667,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 280.57,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 3089.104896,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 22680.69888,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 79.6667,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 280.57,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 36.419,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 37.645,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 37.906,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 39.24,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 108.59,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.129,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 36.263,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 17.128,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 17.604,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.316,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.296,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 36.419,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 37.645,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 37.906,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 39.24,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 108.59,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.129,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 36.263,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 17.128,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 17.604,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.316,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.296,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 16.805,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 17.027,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 17.179,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 17.484,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 236.833,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 16.682,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 7.71,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 8.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.162,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.025,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 16.805,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 17.027,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 17.179,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 17.484,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 236.833,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 16.682,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 7.71,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 8.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.162,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.025,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "resnet50_libtorch_config_5,vgg19_libtorch_config_7": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "resnet50_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "resnet50_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "input",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "output",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "resnet50_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 2,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "resnet50_libtorch_config_5",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "resnet50_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "resnet50_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            },
+                            {
+                                "_model_name": "vgg19_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "vgg19_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 4,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "vgg19_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "vgg19_libtorch_config_7",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "vgg19_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "vgg19_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "resnet50_libtorch_config_5 -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_7 -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_5,vgg19_libtorch_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2680.160256,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23089.64352,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 100.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 274.244,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2680.160256,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23089.64352,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 100.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 274.244,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2680.160256,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23089.64352,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 100.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 274.244,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 28.635,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 30.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 30.354,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 31.183,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 139.873,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.104,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 28.503,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 13.517,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 14.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.163,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 28.635,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 30.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 30.354,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 31.183,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 139.873,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.104,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 28.503,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 13.517,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 14.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.163,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 28.632,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 28.85,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 28.901,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 29.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 279.746,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.178,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 28.425,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 13.041,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 13.196,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.932,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.035,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 28.632,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 28.85,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 28.901,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 29.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 279.746,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.178,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 28.425,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 13.041,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 13.196,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.932,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.035,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "resnet50_libtorch_config_5 -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_7 -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_5,vgg19_libtorch_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2575.302656,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23194.50112,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 93.6667,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 277.039,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2575.302656,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23194.50112,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 93.6667,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 277.039,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2575.302656,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23194.50112,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 93.6667,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 277.039,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 8.091,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 9.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 9.317,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 123.562,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 7.966,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.066,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.922,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.175,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.038,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 8.091,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 9.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 9.317,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 10.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 123.562,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 7.966,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.066,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 6.922,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.175,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.038,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.668,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.837,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.911,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 176.184,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.549,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.638,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.127,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.668,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.837,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.911,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 6.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 176.184,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.549,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.638,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.127,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "resnet50_libtorch_config_0,vgg19_libtorch_config_7": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "resnet50_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "resnet50_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "input",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "output",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "resnet50_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "resnet50_libtorch_config_0",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "resnet50_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "resnet50_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            },
+                            {
+                                "_model_name": "vgg19_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "vgg19_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 4,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "vgg19_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "vgg19_libtorch_config_7",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "vgg19_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "vgg19_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "resnet50_libtorch_config_0 -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_7 -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=8 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_0,vgg19_libtorch_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2401.23904,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23368.564736,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 78.25,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 279.965,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2401.23904,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23368.564736,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 78.25,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 279.965,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2401.23904,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23368.564736,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 78.25,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 279.965,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 26.733,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 26.929,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 26.964,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 27.15,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 74.9411,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 26.567,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 12.549,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 13.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.194,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 26.733,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 26.929,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 26.964,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 27.15,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 74.9411,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 26.567,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 12.549,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 13.034,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.194,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 8,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 26.732,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 26.922,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 26.962,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 27.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 299.764,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 26.571,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 12.132,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 12.557,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.648,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.023,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 26.732,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 26.922,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 26.962,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 27.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 299.764,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.134,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 26.571,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 12.132,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 12.557,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.648,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.023,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "resnet50_libtorch_config_5,vgg19_libtorch_config_4": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "resnet50_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "resnet50_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "input",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "output",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "resnet50_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 2,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "resnet50_libtorch_config_5",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "resnet50_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "resnet50_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            },
+                            {
+                                "_model_name": "vgg19_libtorch",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "vgg19_libtorch",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "format": "FORMAT_NCHW",
+                                                "dims": [
+                                                    "3",
+                                                    "224",
+                                                    "224"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "1000"
+                                                ],
+                                                "labelFilename": "vgg19_labels.txt"
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "vgg19_libtorch_config_4",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "vgg19_libtorch",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "vgg19_libtorch-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "resnet50_libtorch_config_5 -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_4 -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=4 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_5,vgg19_libtorch_config_4",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2602.565632,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23167.238144000003,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 100.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 276.12,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2602.565632,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23167.238144000003,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 100.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 276.12,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2602.565632,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23167.238144000003,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 100.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 276.12,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 21.714,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 23.229,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 23.504,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 23.899,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 184.509,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.11,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 21.575,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 10.059,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 10.574,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.17,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 21.714,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 23.229,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 23.504,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 23.899,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 184.509,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.11,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 21.575,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 10.059,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 10.574,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.17,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.026,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 4,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 16.896,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 17.168,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 17.269,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 17.423,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 236.464,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.133,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 16.737,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 7.479,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 8.023,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.304,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.023,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 16.896,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 17.168,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 17.269,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 17.423,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 236.464,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.133,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 16.737,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 7.479,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 8.023,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.304,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.023,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "resnet50_libtorch_config_5 -m resnet50_libtorch -b 1 -i grpc -f resnet50_libtorch-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000vgg19_libtorch_config_4 -m vgg19_libtorch -b 1 -i grpc -f vgg19_libtorch-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "resnet50_libtorch_config_5,vgg19_libtorch_config_4",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 2575.302656,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 23194.50112,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 98.5,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 276.779,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2575.302656,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23194.50112,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 98.5,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 276.779,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 2575.302656,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 23194.50112,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 98.5,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 276.779,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "resnet50_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.914,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.115,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.348,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 168.905,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.792,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.908,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.12,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.914,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 6.115,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 6.348,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 7.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 168.905,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.792,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.908,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.12,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                },
+                                {
+                                    "_model_config_name": "vgg19_libtorch",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.833,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.921,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.945,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 5.989,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 171.238,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.713,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.036,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.842,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.12,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 5.833,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 5.921,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 5.945,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 5.989,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 171.238,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 5.713,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.036,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 4.842,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.12,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        }
+    },
+    "ResultManager.server_only_data": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+            [
+                "gpu_used_memory",
+                {
+                    "_value": 870.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_free_memory",
+                {
+                    "_value": 24899.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_utilization",
+                {
+                    "_value": 0.0,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ],
+            [
+                "gpu_power_usage",
+                {
+                    "_value": 55.84725,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ]
+        ]
+    },
+    "MetricsManager.gpus": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": {
+            "name": "NVIDIA TITAN RTX",
+            "total_memory": 25387401216
+        }
+    },
+    "ModelManager.model_variant_name_manager": {
+        "_model_config_dicts": {
+            "resnet50_libtorch_config_0": {
+                "name": "resnet50_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "input",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "output",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "resnet50_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_0": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_1": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_2": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_3": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 4,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_4": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "resnet50_libtorch_config_1": {
+                "name": "resnet50_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "input",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "output",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "resnet50_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_5": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 4,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "resnet50_libtorch_config_2": {
+                "name": "resnet50_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "input",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "output",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "resnet50_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "resnet50_libtorch_config_3": {
+                "name": "resnet50_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 4,
+                "input": [
+                    {
+                        "name": "input",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "output",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "resnet50_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "resnet50_libtorch_config_4": {
+                "name": "resnet50_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 4,
+                "input": [
+                    {
+                        "name": "input",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "output",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "resnet50_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "resnet50_libtorch_config_5": {
+                "name": "resnet50_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "input",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "output",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "resnet50_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_6": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 4,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "resnet50_libtorch_config_6": {
+                "name": "resnet50_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "input",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "output",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "resnet50_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "resnet50_libtorch_config_7": {
+                "name": "resnet50_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "input",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "output",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "resnet50_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_7": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 4,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_8": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_9": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 8,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "vgg19_libtorch_config_10": {
+                "name": "vgg19_libtorch",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "format": "FORMAT_NCHW",
+                        "dims": [
+                            "3",
+                            "224",
+                            "224"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "1000"
+                        ],
+                        "label_filename": "vgg19_labels.txt"
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            }
+        },
+        "_model_name_index": {
+            "resnet50_libtorch": 7,
+            "vgg19_libtorch": 10
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/c++/perf_analyzer/tests/common/multi-model-ckpt/README b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/README
new file mode 100644
index 000000000..933296cff
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/README
@@ -0,0 +1,19 @@
+# Copyright (c) 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+Settings used to create this checkpoint:
+
+--profile-models resnet50_libtorch,vgg19_libtorch
+--run-config-profile-models-concurrently-enable
+--run-config-search-mode quick
\ No newline at end of file
diff --git a/src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-model-gpu.csv b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-model-gpu.csv
new file mode 100644
index 000000000..d8cf2bf36
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-model-gpu.csv
@@ -0,0 +1,14 @@
+Model,GPU UUID,Batch,Concurrency,Model Config Path,Instance Group,Satisfies Constraints,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","4,4","resnet50_libtorch_config_5,vgg19_libtorch_config_4","2:GPU,1:GPU",Yes,2602.6,100.0,276.1
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","1,1","resnet50_libtorch_config_5,vgg19_libtorch_config_4","2:GPU,1:GPU",Yes,2575.3,98.5,276.8
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","4,8","resnet50_libtorch_config_5,vgg19_libtorch_config_7","2:GPU,1:GPU",Yes,2680.2,100.0,274.2
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","1,1","resnet50_libtorch_config_5,vgg19_libtorch_config_7","2:GPU,1:GPU",Yes,2575.3,93.7,277.0
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","1,1","resnet50_libtorch_config_default,vgg19_libtorch_config_default","1:GPU,1:GPU",Yes,2296.4,99.7,273.8
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","4,4","resnet50_libtorch_config_default,vgg19_libtorch_config_default","1:GPU,1:GPU",Yes,2296.4,50.5,178.0
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","2,2","resnet50_libtorch_config_default,vgg19_libtorch_config_default","1:GPU,1:GPU",Yes,2296.4,99.5,277.9
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","8,8","resnet50_libtorch_config_default,vgg19_libtorch_config_default","1:GPU,1:GPU",Yes,2296.4,60.3,216.1
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","2,4","resnet50_libtorch_config_0,vgg19_libtorch_config_4","1:GPU,1:GPU",Yes,2323.6,99.3,275.4
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","2,2","resnet50_libtorch_config_0,vgg19_libtorch_config_0","1:GPU,1:GPU",Yes,2296.4,97.7,269.3
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","4,4","resnet50_libtorch_config_7,vgg19_libtorch_config_1","1:GPU,2:GPU",Yes,3089.1,79.7,280.6
+"resnet50_libtorch,vgg19_libtorch",GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,"1,1","2,8","resnet50_libtorch_config_0,vgg19_libtorch_config_7","1:GPU,1:GPU",Yes,2401.2,78.2,280.0
+
diff --git a/src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-model-inference.csv b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-model-inference.csv
new file mode 100644
index 000000000..8373a448f
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-model-inference.csv
@@ -0,0 +1,14 @@
+Model,Batch,Concurrency,Model Config Path,Instance Group,Max Batch Size,Satisfies Constraints,Throughput (infer/sec),p99 Latency (ms)
+"resnet50_libtorch,vgg19_libtorch","1,1","4,4","resnet50_libtorch_config_5,vgg19_libtorch_config_4","2:GPU,1:GPU","1,2",Yes,"421.0, [184.5,236.5]","20.7, [23.9,17.4]"
+"resnet50_libtorch,vgg19_libtorch","1,1","1,1","resnet50_libtorch_config_5,vgg19_libtorch_config_4","2:GPU,1:GPU","1,2",Yes,"340.1, [168.9,171.2]","6.5, [7.0,6.0]"
+"resnet50_libtorch,vgg19_libtorch","1,1","4,8","resnet50_libtorch_config_5,vgg19_libtorch_config_7","2:GPU,1:GPU","1,4",Yes,"419.6, [139.9,279.7]","30.1, [31.2,29.0]"
+"resnet50_libtorch,vgg19_libtorch","1,1","1,1","resnet50_libtorch_config_5,vgg19_libtorch_config_7","2:GPU,1:GPU","1,4",Yes,"299.7, [123.6,176.2]","8.0, [10.0,6.0]"
+"resnet50_libtorch,vgg19_libtorch","1,1","1,1","resnet50_libtorch_config_default,vgg19_libtorch_config_default","1:GPU,1:GPU","128,128",Yes,"344.1, [167.9,176.2]","6.8, [7.7,5.9]"
+"resnet50_libtorch,vgg19_libtorch","1,1","4,4","resnet50_libtorch_config_default,vgg19_libtorch_config_default","1:GPU,1:GPU","128,128",Yes,"338.8, [139.2,199.5]","25.6, [30.4,20.7]"
+"resnet50_libtorch,vgg19_libtorch","1,1","2,2","resnet50_libtorch_config_default,vgg19_libtorch_config_default","1:GPU,1:GPU","128,128",Yes,"339.1, [138.6,200.5]","13.0, [15.6,10.3]"
+"resnet50_libtorch,vgg19_libtorch","1,1","8,8","resnet50_libtorch_config_default,vgg19_libtorch_config_default","1:GPU,1:GPU","128,128",Yes,"338.1, [138.5,199.5]","50.7, [60.7,40.7]"
+"resnet50_libtorch,vgg19_libtorch","1,1","2,4","resnet50_libtorch_config_0,vgg19_libtorch_config_4","1:GPU,1:GPU","1,2",Yes,"368.4, [108.6,259.8]","17.9, [19.9,15.8]"
+"resnet50_libtorch,vgg19_libtorch","1,1","2,2","resnet50_libtorch_config_0,vgg19_libtorch_config_0","1:GPU,1:GPU","1,1",Yes,"341.4, [139.6,201.8]","12.9, [15.5,10.3]"
+"resnet50_libtorch,vgg19_libtorch","1,1","4,4","resnet50_libtorch_config_7,vgg19_libtorch_config_1","1:GPU,2:GPU","2,1",Yes,"345.4, [108.6,236.8]","28.4, [39.2,17.5]"
+"resnet50_libtorch,vgg19_libtorch","1,1","2,8","resnet50_libtorch_config_0,vgg19_libtorch_config_7","1:GPU,1:GPU","1,4",Yes,"374.7, [74.9,299.8]","27.1, [27.1,27.1]"
+
diff --git a/src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-server-only.csv b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-server-only.csv
new file mode 100644
index 000000000..25a6071cf
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/golden-metrics-server-only.csv
@@ -0,0 +1,3 @@
+Model,GPU UUID,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+triton-server,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,870.0,0.0,55.8
+
diff --git a/src/c++/perf_analyzer/tests/common/multi-model-ckpt/plot_manager.json b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/plot_manager.json
new file mode 100644
index 000000000..7ad1567c5
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/multi-model-ckpt/plot_manager.json
@@ -0,0 +1 @@
+{"_simple_plots": {"resnet50_libtorch,vgg19_libtorch": {"throughput_v_latency": {"_name": "throughput_v_latency", "_title": "Throughput vs. Latency", "_x_axis": "perf_latency_p99", "_y_axis": "perf_throughput", "_monotonic": true, "_data": {"resnet50_libtorch_config_5,vgg19_libtorch_config_4": {"x_data": [20.661, 6.497], "y_data": [420.97299999999996, 340.14300000000003]}, "resnet50_libtorch_config_5,vgg19_libtorch_config_7": {"x_data": [30.093, 8.0365], "y_data": [419.61899999999997, 299.746]}, "resnet50_libtorch_config_default,vgg19_libtorch_config_default": {"x_data": [6.7745, 25.5655, 12.9815, 50.6635], "y_data": [344.119, 338.762, 339.052, 338.05600000000004]}}}, "gpu_mem_v_latency": {"_name": "gpu_mem_v_latency", "_title": "GPU Memory vs. Latency", "_x_axis": "perf_latency_p99", "_y_axis": "gpu_used_memory", "_monotonic": false, "_data": {"resnet50_libtorch_config_5,vgg19_libtorch_config_4": {"x_data": [20.661, 6.497], "y_data": [2602.565632, 2575.302656]}, "resnet50_libtorch_config_5,vgg19_libtorch_config_7": {"x_data": [30.093, 8.0365], "y_data": [2680.160256, 2575.302656]}, "resnet50_libtorch_config_default,vgg19_libtorch_config_default": {"x_data": [6.7745, 25.5655, 12.9815, 50.6635], "y_data": [2296.38144, 2296.38144, 2296.38144, 2296.38144]}}}}}}
\ No newline at end of file
diff --git a/src/c++/perf_analyzer/tests/common/request-rate-ckpt/0.ckpt b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/0.ckpt
new file mode 100644
index 000000000..0f224d4c5
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/0.ckpt
@@ -0,0 +1,11428 @@
+{
+    "ResultManager.results": {
+        "_results": {
+            "add_sub": {
+                "add_sub_config_default": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 8,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ]
+                                    },
+                                    "variant_name": "add_sub_config_default",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_default -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.4722,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.4722,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.4722,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.061,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.225,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.36,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.427,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.079,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.073,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.155,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.104,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.061,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.225,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.36,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.427,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.079,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.073,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.155,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.104,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_default -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.4765,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.4765,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.4765,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.02,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.122,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.258,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.36,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9887,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.975,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.07,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.144,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.093,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.02,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.122,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.258,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.36,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9887,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.975,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.07,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.144,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.093,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_default -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.5077,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5077,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5077,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.122,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.282,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.417,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.532,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1578,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.078,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.167,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.107,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.122,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.282,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.417,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.532,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1578,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.078,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.167,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.107,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_default -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.5913,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5913,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5913,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.073,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.283,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.339,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.392,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9755,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.025,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.162,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.073,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.283,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.339,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.392,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9755,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.025,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.162,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_0": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_0",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_0 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.691,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.691,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.691,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.237,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.369,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.506,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.146,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.237,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.369,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.506,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.146,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_0 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7322,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7322,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7322,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.07,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.19,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.336,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.466,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9878,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.024,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.091,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.07,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.19,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.336,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.466,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9878,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.024,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.143,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.091,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_1": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_1",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_1 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7208,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7208,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7208,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.12,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.225,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.307,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.508,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0794,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.133,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.12,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.225,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.307,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.508,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0794,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.133,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_1 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7795,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7795,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7795,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.094,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.212,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.294,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.452,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9889,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.049,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.144,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.142,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.091,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.094,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.212,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.294,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.452,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9889,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.049,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.144,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.142,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.091,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_1 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7322,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7322,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7322,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.118,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.263,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.329,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.596,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.158,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.071,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.141,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.155,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.093,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.118,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.263,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.329,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.596,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.158,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.071,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.141,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.155,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.093,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_1 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7118,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7118,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7118,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.199,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.49,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9739,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.051,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.139,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.152,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.199,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.332,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.49,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9739,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.051,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.139,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.152,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_2": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 2,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_2",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_2 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.8955,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8955,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8955,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.132,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.251,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.357,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.489,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.085,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.139,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.153,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.102,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.132,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.251,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.357,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.489,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.085,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.139,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.153,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.102,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_2 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7785,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7785,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7785,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.127,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.259,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.384,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.527,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9868,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.078,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.163,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.107,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.127,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.259,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.384,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.527,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9868,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.078,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.163,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.107,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_3": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 2,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_3",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_3 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9191,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9191,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9191,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.116,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.235,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.442,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0792,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.068,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.116,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.235,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.442,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0792,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.068,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_3 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9508,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9508,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9508,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.193,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.45,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9884,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.059,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.146,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.193,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.45,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9884,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.059,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.146,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_3 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9373,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9373,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9373,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.122,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.231,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.267,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.402,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1575,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.155,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.122,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.231,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.267,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.402,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1575,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.138,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.155,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_3 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.8725,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8725,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.8725,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.127,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.339,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.447,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.518,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9765,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.081,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.16,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.127,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.339,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.447,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.518,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9765,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.081,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.16,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_4": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 3,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_4",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_4 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_4",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 876.609536,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24893.19424,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9247,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9247,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9247,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.21,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.263,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.51,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0798,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.052,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.21,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.263,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.51,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0798,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.052,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_4 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_4",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 876.609536,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24893.19424,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9072,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9072,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9072,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.212,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.27,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.461,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9889,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.054,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.145,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.212,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.27,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.461,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9889,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.054,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.136,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.145,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_5": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 3,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_5",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_5 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 876.609536,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24893.19424,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9405,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9405,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9405,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.212,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.284,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.384,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.053,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.212,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.284,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.384,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.053,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_5 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 876.609536,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24893.19424,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9038,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9038,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9038,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.113,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.22,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.328,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.498,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9885,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.065,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.141,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.113,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.22,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.328,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.498,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9885,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.065,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.141,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.147,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_5 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 876.609536,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24893.19424,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 57.0292,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0292,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0292,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.145,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.28,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.325,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.498,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.162,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.145,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.28,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.325,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.498,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.162,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_5 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_5",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 876.609536,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24893.19424,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 57.0632,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0632,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 876.609536,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24893.19424,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0632,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.118,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.23,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.349,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.445,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9786,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.154,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.093,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.118,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.23,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.349,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.445,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9786,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.072,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.154,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.093,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_6": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 4,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_6",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_6 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 878.706688,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24891.097088000002,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 57.0973,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0973,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0973,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.219,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.314,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.547,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.219,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.314,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.547,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_6 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_6",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 878.706688,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24891.097088000002,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 57.1385,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.1385,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.1385,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.187,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.258,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.475,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9883,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.057,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.145,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.187,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.258,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.475,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9883,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.057,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.145,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.094,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_7": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 4,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_7",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_7 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 878.706688,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24891.097088000002,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 57.0168,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0168,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0168,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.125,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.213,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.296,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.451,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0792,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.079,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.125,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.213,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.296,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.451,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0792,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.079,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.149,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_7 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 878.706688,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24891.097088000002,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 57.0233,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0233,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0233,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.34,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.433,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.517,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9895,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.075,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.15,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.102,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.34,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.433,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.517,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9895,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.075,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.15,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.102,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_7 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 878.706688,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24891.097088000002,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 57.0362,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0362,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0362,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.126,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.243,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.3,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.522,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1576,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.079,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.159,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.126,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.243,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.3,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.522,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1576,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.079,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.137,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.159,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_7 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_7",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 878.706688,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24891.097088000002,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 57.125,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.125,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 878.706688,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24891.097088000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.125,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.104,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.18,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.288,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.441,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9766,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.059,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.151,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.104,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.18,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.288,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.441,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9766,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.059,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.151,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.092,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_8": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 5,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_8",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_8 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_8",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 880.80384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24888.999936,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9587,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9587,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9587,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.132,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.285,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.38,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.472,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.083,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.152,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.104,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.132,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.285,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.38,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.472,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0791,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.083,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.14,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.152,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.104,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_8 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_8",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 880.80384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24888.999936,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.942,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.942,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.942,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.223,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.353,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.445,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9868,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.133,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.121,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.223,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.353,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.445,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9868,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.074,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.133,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.1,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.097,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_9": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 5,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_9",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_9 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_9",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 880.80384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24888.999936,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9757,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9757,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9757,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.138,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.269,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.345,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.516,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0795,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.091,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.142,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.153,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.138,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.269,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.345,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.516,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 16.0795,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.091,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.142,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.153,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_9 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_9",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 880.80384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24888.999936,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.9743,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9743,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.9743,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.2,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.285,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.488,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9887,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.052,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.128,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.097,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.2,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.285,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.488,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 31.9887,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.012,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.052,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.128,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.148,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.096,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_9 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=16 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_9",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 880.80384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24888.999936,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 57.0045,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0045,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 57.0045,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 16
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.151,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.359,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.511,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.102,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.165,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.151,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.302,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.359,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.511,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 32.1583,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.014,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.102,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.165,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.103,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_9 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --request-rate-range=32 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_9",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 880.80384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24888.999936,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 0.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.904,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.904,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 880.80384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24888.999936,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 0.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.904,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": null,
+                                        "request-rate-range": 32
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.112,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.224,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.338,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.459,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9746,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.066,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.156,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 1.112,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 1.224,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 1.338,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 1.459,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 63.9746,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.013,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 1.066,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.135,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.156,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.095,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        }
+    },
+    "ResultManager.server_only_data": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+            [
+                "gpu_used_memory",
+                {
+                    "_value": 870.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_free_memory",
+                {
+                    "_value": 24899.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_utilization",
+                {
+                    "_value": 0.0,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ],
+            [
+                "gpu_power_usage",
+                {
+                    "_value": 56.12249999999999,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ]
+        ]
+    },
+    "MetricsManager.gpus": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": {
+            "name": "NVIDIA TITAN RTX",
+            "total_memory": 25387401216
+        }
+    },
+    "ModelManager.model_variant_name_manager": {
+        "_model_config_dicts": {
+            "add_sub_config_0": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_1": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_2": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_3": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_4": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_5": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 3,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_6": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 4,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_7": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 4,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_8": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 5,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_9": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 5,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            }
+        },
+        "_model_name_index": {
+            "add_sub": 9
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/c++/perf_analyzer/tests/common/request-rate-ckpt/README b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/README
new file mode 100644
index 000000000..3a2377082
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/README
@@ -0,0 +1,21 @@
+# Copyright (c) 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+Settings used to create this checkpoint:
+
+--profile-models add_sub
+--request-rate 16,32
+--run-config-search-max-instance-count 2
+--run-config-search-max-model-batch-size 2
+--request-rate-search-enable
diff --git a/src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-model-gpu.csv b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-model-gpu.csv
new file mode 100644
index 000000000..fd74ba39f
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-model-gpu.csv
@@ -0,0 +1,36 @@
+Model,GPU UUID,Batch,Request Rate,Model Config Path,Instance Group,Satisfies Constraints,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,32,add_sub_config_5,3:GPU,Yes,876.6,0.0,57.1
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,16,add_sub_config_5,3:GPU,Yes,876.6,0.0,57.0
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_5,3:GPU,Yes,876.6,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_5,3:GPU,Yes,876.6,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,32,add_sub_config_7,4:GPU,Yes,878.7,0.0,57.1
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,16,add_sub_config_7,4:GPU,Yes,878.7,0.0,57.0
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_7,4:GPU,Yes,878.7,0.0,57.0
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_7,4:GPU,Yes,878.7,0.0,57.0
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,32,add_sub_config_3,2:GPU,Yes,874.5,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,16,add_sub_config_3,2:GPU,Yes,874.5,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_3,2:GPU,Yes,874.5,0.0,57.0
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_3,2:GPU,Yes,874.5,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,32,add_sub_config_default,1:GPU,Yes,872.4,0.0,56.6
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,16,add_sub_config_default,1:GPU,Yes,872.4,0.0,56.5
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_default,1:GPU,Yes,872.4,0.0,56.5
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_default,1:GPU,Yes,872.4,0.0,56.5
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,32,add_sub_config_9,5:GPU,Yes,880.8,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,16,add_sub_config_9,5:GPU,Yes,880.8,0.0,57.0
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_9,5:GPU,Yes,880.8,0.0,57.0
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_9,5:GPU,Yes,880.8,0.0,57.0
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,32,add_sub_config_1,1:GPU,Yes,872.4,0.0,56.7
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,16,add_sub_config_1,1:GPU,Yes,872.4,0.0,56.7
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_1,1:GPU,Yes,872.4,0.0,56.8
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_1,1:GPU,Yes,872.4,0.0,56.7
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_4,3:GPU,Yes,876.6,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_4,3:GPU,Yes,876.6,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_6,4:GPU,Yes,878.7,0.0,57.1
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_6,4:GPU,Yes,878.7,0.0,57.1
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_0,1:GPU,Yes,872.4,0.0,56.7
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_0,1:GPU,Yes,872.4,0.0,56.7
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_2,2:GPU,Yes,874.5,0.0,56.8
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_2,2:GPU,Yes,874.5,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,32,add_sub_config_8,5:GPU,Yes,880.8,0.0,56.9
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,16,add_sub_config_8,5:GPU,Yes,880.8,0.0,57.0
+
diff --git a/src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-model-inference.csv b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-model-inference.csv
new file mode 100644
index 000000000..1f7daaf25
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-model-inference.csv
@@ -0,0 +1,36 @@
+Model,Batch,Request Rate,Model Config Path,Instance Group,Max Batch Size,Satisfies Constraints,Throughput (infer/sec),p99 Latency (ms)
+add_sub,2,32,add_sub_config_5,3:GPU,2,Yes,64.0,1.4
+add_sub,2,16,add_sub_config_5,3:GPU,2,Yes,32.2,1.5
+add_sub,1,32,add_sub_config_5,3:GPU,2,Yes,32.0,1.5
+add_sub,1,16,add_sub_config_5,3:GPU,2,Yes,16.1,1.4
+add_sub,2,32,add_sub_config_7,4:GPU,2,Yes,64.0,1.4
+add_sub,2,16,add_sub_config_7,4:GPU,2,Yes,32.2,1.5
+add_sub,1,32,add_sub_config_7,4:GPU,2,Yes,32.0,1.5
+add_sub,1,16,add_sub_config_7,4:GPU,2,Yes,16.1,1.5
+add_sub,2,32,add_sub_config_3,2:GPU,2,Yes,64.0,1.5
+add_sub,2,16,add_sub_config_3,2:GPU,2,Yes,32.2,1.4
+add_sub,1,32,add_sub_config_3,2:GPU,2,Yes,32.0,1.4
+add_sub,1,16,add_sub_config_3,2:GPU,2,Yes,16.1,1.4
+add_sub,2,32,add_sub_config_default,1:GPU,8,Yes,64.0,1.4
+add_sub,2,16,add_sub_config_default,1:GPU,8,Yes,32.2,1.5
+add_sub,1,32,add_sub_config_default,1:GPU,8,Yes,32.0,1.4
+add_sub,1,16,add_sub_config_default,1:GPU,8,Yes,16.1,1.4
+add_sub,2,32,add_sub_config_9,5:GPU,2,Yes,64.0,1.5
+add_sub,2,16,add_sub_config_9,5:GPU,2,Yes,32.2,1.5
+add_sub,1,32,add_sub_config_9,5:GPU,2,Yes,32.0,1.5
+add_sub,1,16,add_sub_config_9,5:GPU,2,Yes,16.1,1.5
+add_sub,2,32,add_sub_config_1,1:GPU,2,Yes,64.0,1.5
+add_sub,2,16,add_sub_config_1,1:GPU,2,Yes,32.2,1.6
+add_sub,1,32,add_sub_config_1,1:GPU,2,Yes,32.0,1.5
+add_sub,1,16,add_sub_config_1,1:GPU,2,Yes,16.1,1.5
+add_sub,1,32,add_sub_config_4,3:GPU,1,Yes,32.0,1.5
+add_sub,1,16,add_sub_config_4,3:GPU,1,Yes,16.1,1.5
+add_sub,1,32,add_sub_config_6,4:GPU,1,Yes,32.0,1.5
+add_sub,1,16,add_sub_config_6,4:GPU,1,Yes,16.1,1.5
+add_sub,1,32,add_sub_config_0,1:GPU,1,Yes,32.0,1.5
+add_sub,1,16,add_sub_config_0,1:GPU,1,Yes,16.1,1.5
+add_sub,1,32,add_sub_config_2,2:GPU,1,Yes,32.0,1.5
+add_sub,1,16,add_sub_config_2,2:GPU,1,Yes,16.1,1.5
+add_sub,1,32,add_sub_config_8,5:GPU,1,Yes,32.0,1.4
+add_sub,1,16,add_sub_config_8,5:GPU,1,Yes,16.1,1.5
+
diff --git a/src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-server-only.csv b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-server-only.csv
new file mode 100644
index 000000000..4daa8fc3d
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/request-rate-ckpt/golden-metrics-server-only.csv
@@ -0,0 +1,3 @@
+Model,GPU UUID,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+triton-server,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,870.0,0.0,56.1
+
diff --git a/src/c++/perf_analyzer/tests/common/single-model-ckpt/0.ckpt b/src/c++/perf_analyzer/tests/common/single-model-ckpt/0.ckpt
new file mode 100644
index 000000000..a3d9e0489
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/single-model-ckpt/0.ckpt
@@ -0,0 +1,5350 @@
+{
+    "ResultManager.results": {
+        "_results": {
+            "add_sub": {
+                "add_sub_config_default": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 8,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ]
+                                    },
+                                    "variant_name": "add_sub_config_default",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_default -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 2.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.1825,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 2.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.1825,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 2.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.1825,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.397,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.463,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.478,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.539,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2508.04,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.382,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.064,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.042,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.397,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.463,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.478,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.539,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2508.04,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.382,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.064,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.042,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_default -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 4.25,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.3008,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 4.25,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.3008,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 4.25,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.3008,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.322,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.401,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.465,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.581,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 6176.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.309,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.322,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.401,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.465,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.581,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 6176.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.309,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_default -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 3.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.181,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.181,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.181,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.409,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.497,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.56,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.658,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 4876.12,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.393,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.07,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.037,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.409,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.497,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.56,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.658,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 4876.12,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.393,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.07,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.037,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_default -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_default",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 5.5,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.3928,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 5.5,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.3928,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 5.5,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.3928,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.301,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.344,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.384,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.524,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 13205.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.288,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.301,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.344,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.384,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.524,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 13205.0,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.288,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.029,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.03,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_0": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_0",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_0 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 2.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.391,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 2.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.391,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 2.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.391,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.427,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.481,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.505,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.564,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2334.71,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.412,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.069,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.037,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.427,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.481,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.505,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.564,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2334.71,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.412,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.039,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.069,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.037,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_0 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_0",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 4.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.5265,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 4.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5265,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 4.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5265,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.326,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.383,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.455,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.599,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 6114.93,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.312,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.326,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.383,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.455,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.599,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 6114.93,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.312,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.031,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.019,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_1": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 1,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_1",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_1 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 1.7500000000000002,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.417,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 1.7500000000000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.417,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 1.7500000000000002,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.417,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.456,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.53,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.546,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.599,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2185.97,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.44,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.041,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.071,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.04,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.456,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.53,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.546,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.599,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 2185.97,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.44,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.041,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.071,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.04,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_1 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 4.75,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6202,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 4.75,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6202,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 4.75,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6202,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.326,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.389,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.474,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.607,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 6109.95,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.313,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.048,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.326,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.389,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.474,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.607,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 6109.95,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.313,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.048,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.033,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.02,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.022,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_1 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 3.5000000000000004,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.479,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.5000000000000004,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.479,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.5000000000000004,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.479,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.415,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.477,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.511,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.587,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 4804.46,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.4,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.038,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.069,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.041,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.415,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.477,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.511,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.587,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 4804.46,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.4,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.038,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.069,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.035,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.041,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_1 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_1",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 872.415232,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24897.388544,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 6.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.5925,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 6.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5925,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 872.415232,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24897.388544,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 6.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5925,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.318,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.36,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.421,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.596,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 12528.9,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.304,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.042,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.318,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.36,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.421,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.596,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 12528.9,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.003,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.304,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.042,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.032,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.018,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.021,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_2": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 1,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 2,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_2",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": null,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_2 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 3.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.5757,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5757,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.5757,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.52,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.655,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.718,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.789,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1915.57,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.503,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.089,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.052,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.056,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.52,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.655,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.718,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.789,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1915.57,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.503,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.047,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.089,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.052,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.056,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_2 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_2",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 2.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6993,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 2.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6993,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 2.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6993,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.593,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.721,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.764,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.864,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3363.11,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.573,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.055,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.061,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.593,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.721,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.764,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.864,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3363.11,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.573,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.044,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.098,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.055,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.061,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ],
+                "add_sub_config_3": [
+                    {
+                        "_triton_env": {},
+                        "_model_run_configs": [
+                            {
+                                "_model_name": "add_sub",
+                                "_model_config_variant": {
+                                    "model_config": {
+                                        "name": "add_sub",
+                                        "platform": "pytorch_libtorch",
+                                        "maxBatchSize": 2,
+                                        "input": [
+                                            {
+                                                "name": "INPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            },
+                                            {
+                                                "name": "INPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "output": [
+                                            {
+                                                "name": "OUTPUT__0",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ],
+                                                "labelFilename": "output0_labels.txt"
+                                            },
+                                            {
+                                                "name": "OUTPUT__1",
+                                                "dataType": "TYPE_FP32",
+                                                "dims": [
+                                                    "16"
+                                                ]
+                                            }
+                                        ],
+                                        "instanceGroup": [
+                                            {
+                                                "count": 2,
+                                                "kind": "KIND_GPU"
+                                            }
+                                        ],
+                                        "dynamicBatching": {}
+                                    },
+                                    "variant_name": "add_sub_config_3",
+                                    "cpu_only": false
+                                },
+                                "_perf_config": {
+                                    "_args": {
+                                        "service-kind": null,
+                                        "model-signature-name": null,
+                                        "async": null,
+                                        "sync": null,
+                                        "measurement-interval": null,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null,
+                                        "request-distribution": null,
+                                        "request-intervals": null,
+                                        "binary-search": null,
+                                        "num-of-sequences": null,
+                                        "latency-threshold": null,
+                                        "max-threads": null,
+                                        "stability-percentage": null,
+                                        "max-trials": null,
+                                        "percentile": null,
+                                        "input-data": null,
+                                        "shared-memory": null,
+                                        "output-shared-memory-size": null,
+                                        "shape": null,
+                                        "sequence-length": null,
+                                        "sequence-id-range": null,
+                                        "string-length": null,
+                                        "string-data": null,
+                                        "measurement-mode": "count_windows",
+                                        "measurement-request-count": 100,
+                                        "streaming": null,
+                                        "grpc-compression-algorithm": null,
+                                        "triton-server-directory": null,
+                                        "model-repository": null,
+                                        "ssl-grpc-use-ssl": null,
+                                        "ssl-grpc-root-certifications-file": null,
+                                        "ssl-grpc-private-key-file": null,
+                                        "ssl-grpc-certificate-chain-file": null,
+                                        "ssl-https-verify-peer": null,
+                                        "ssl-https-verify-host": null,
+                                        "ssl-https-ca-certificates-file": null,
+                                        "ssl-https-client-certificate-type": null,
+                                        "ssl-https-client-certificate-file": null,
+                                        "ssl-https-private-key-type": null,
+                                        "ssl-https-private-key-file": null,
+                                        "collect-metrics": "True",
+                                        "metrics-url": "http://localhost:8002/metrics",
+                                        "metrics-interval": 1000,
+                                        "bls-composing-models": null
+                                    },
+                                    "_options": {
+                                        "-m": "add_sub",
+                                        "-x": null,
+                                        "-b": 1,
+                                        "-u": "localhost:8001",
+                                        "-i": "grpc",
+                                        "-f": "add_sub-results.csv",
+                                        "-H": null
+                                    },
+                                    "_verbose": {
+                                        "-v": null,
+                                        "-v -v": null,
+                                        "--verbose-csv": "--verbose-csv"
+                                    },
+                                    "_input_to_options": {
+                                        "model-name": "-m",
+                                        "model-version": "-x",
+                                        "batch-size": "-b",
+                                        "url": "-u",
+                                        "protocol": "-i",
+                                        "latency-report-file": "-f",
+                                        "http-header": "-H"
+                                    },
+                                    "_input_to_verbose": {
+                                        "verbose": "-v",
+                                        "extra-verbose": "-v -v",
+                                        "verbose-csv": "--verbose-csv"
+                                    },
+                                    "_additive_args": {
+                                        "input-data": null,
+                                        "shape": null
+                                    }
+                                },
+                                "_composing_config_variants": []
+                            }
+                        ]
+                    },
+                    {
+                        "add_sub_config_3 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 1.6666699999999999,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.6563,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 1.6666699999999999,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6563,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 1.6666699999999999,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.6563,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.569,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.74,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.767,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.839,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1752.08,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.552,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.056,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.061,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.064,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.569,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.74,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.767,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.839,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 1752.08,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.552,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.056,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.099,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.061,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.064,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_3 -m add_sub -b 1 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 3.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.621,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.621,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.621,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 1,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.609,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.712,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.754,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.838,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3270.69,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.589,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.045,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.057,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.063,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.609,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.712,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.754,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.838,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3270.69,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.005,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.589,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.045,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.101,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.057,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.063,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_3 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=1 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 3.0,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.7673,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7673,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 3.0,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.7673,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": 1,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.5,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.658,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.709,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.803,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3988.65,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.483,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.09,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.048,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.5,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.658,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.709,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.803,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 3988.65,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.483,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.046,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.09,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.048,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.054,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        },
+                        "add_sub_config_3 -m add_sub -b 2 -i grpc -f add_sub-results.csv --verbose-csv --concurrency-range=2 --measurement-mode=count_windows --collect-metrics --metrics-url=http://localhost:8002/metrics --metrics-interval=1000": {
+                            "_model_variants_name": "add_sub_config_3",
+                            "_gpu_data": {
+                                "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+                                    [
+                                        "gpu_used_memory",
+                                        {
+                                            "_value": 874.512384,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_free_memory",
+                                        {
+                                            "_value": 24895.291392,
+                                            "_timestamp": 0,
+                                            "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                                        }
+                                    ],
+                                    [
+                                        "gpu_utilization",
+                                        {
+                                            "_value": 2.75,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ],
+                                    [
+                                        "gpu_power_usage",
+                                        {
+                                            "_value": 56.815,
+                                            "_timestamp": 0,
+                                            "_device_uuid": null
+                                        }
+                                    ]
+                                ]
+                            },
+                            "_avg_gpu_data": [
+                                [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 2.75,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.815,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            ],
+                            "_avg_gpu_data_from_tag": {
+                                "gpu_used_memory": [
+                                    "gpu_used_memory",
+                                    {
+                                        "_value": 874.512384,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_free_memory": [
+                                    "gpu_free_memory",
+                                    {
+                                        "_value": 24895.291392,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_utilization": [
+                                    "gpu_utilization",
+                                    {
+                                        "_value": 2.75,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ],
+                                "gpu_power_usage": [
+                                    "gpu_power_usage",
+                                    {
+                                        "_value": 56.815,
+                                        "_timestamp": 0,
+                                        "_device_uuid": null
+                                    }
+                                ]
+                            },
+                            "_model_config_measurements": [
+                                {
+                                    "_model_config_name": "add_sub",
+                                    "_model_specific_pa_params": {
+                                        "batch-size": 2,
+                                        "concurrency-range": 2,
+                                        "request-rate-range": null
+                                    },
+                                    "_non_gpu_data": [
+                                        [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.572,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.677,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.715,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.81,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 6967.92,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.554,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.102,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.059,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    ],
+                                    "_non_gpu_data_from_tag": {
+                                        "perf_latency_avg": [
+                                            "perf_latency_avg",
+                                            {
+                                                "_value": 0.572,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p90": [
+                                            "perf_latency_p90",
+                                            {
+                                                "_value": 0.677,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p95": [
+                                            "perf_latency_p95",
+                                            {
+                                                "_value": 0.715,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_latency_p99": [
+                                            "perf_latency_p99",
+                                            {
+                                                "_value": 0.81,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_throughput": [
+                                            "perf_throughput",
+                                            {
+                                                "_value": 6967.92,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_send_recv": [
+                                            "perf_client_send_recv",
+                                            {
+                                                "_value": 0.004,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_client_response_wait": [
+                                            "perf_client_response_wait",
+                                            {
+                                                "_value": 0.554,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_queue": [
+                                            "perf_server_queue",
+                                            {
+                                                "_value": 0.043,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_infer": [
+                                            "perf_server_compute_infer",
+                                            {
+                                                "_value": 0.102,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_input": [
+                                            "perf_server_compute_input",
+                                            {
+                                                "_value": 0.053,
+                                                "_timestamp": 0
+                                            }
+                                        ],
+                                        "perf_server_compute_output": [
+                                            "perf_server_compute_output",
+                                            {
+                                                "_value": 0.059,
+                                                "_timestamp": 0
+                                            }
+                                        ]
+                                    }
+                                }
+                            ]
+                        }
+                    }
+                ]
+            }
+        }
+    },
+    "ResultManager.server_only_data": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": [
+            [
+                "gpu_used_memory",
+                {
+                    "_value": 870.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_free_memory",
+                {
+                    "_value": 24899.0,
+                    "_timestamp": 0,
+                    "_device_uuid": "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0"
+                }
+            ],
+            [
+                "gpu_utilization",
+                {
+                    "_value": 0.0,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ],
+            [
+                "gpu_power_usage",
+                {
+                    "_value": 55.95925,
+                    "_timestamp": 0,
+                    "_device_uuid": null
+                }
+            ]
+        ]
+    },
+    "MetricsManager.gpus": {
+        "GPU-8557549f-9c89-4384-8bd6-1fd823c342e0": {
+            "name": "NVIDIA TITAN RTX",
+            "total_memory": 25387401216
+        }
+    },
+    "ModelManager.model_variant_name_manager": {
+        "_model_config_dicts": {
+            "add_sub_config_0": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_1": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 1,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_2": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 1,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            },
+            "add_sub_config_3": {
+                "name": "add_sub",
+                "platform": "pytorch_libtorch",
+                "max_batch_size": 2,
+                "input": [
+                    {
+                        "name": "INPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    },
+                    {
+                        "name": "INPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "output": [
+                    {
+                        "name": "OUTPUT__0",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ],
+                        "label_filename": "output0_labels.txt"
+                    },
+                    {
+                        "name": "OUTPUT__1",
+                        "data_type": "TYPE_FP32",
+                        "dims": [
+                            "16"
+                        ]
+                    }
+                ],
+                "instance_group": [
+                    {
+                        "count": 2,
+                        "kind": "KIND_GPU"
+                    }
+                ],
+                "dynamic_batching": {}
+            }
+        },
+        "_model_name_index": {
+            "add_sub": 3
+        }
+    }
+}
\ No newline at end of file
diff --git a/src/c++/perf_analyzer/tests/common/single-model-ckpt/README b/src/c++/perf_analyzer/tests/common/single-model-ckpt/README
new file mode 100644
index 000000000..897c8353c
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/single-model-ckpt/README
@@ -0,0 +1,22 @@
+# Copyright (c) 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+Settings used to create this checkpoint:
+
+--profile-models add_sub
+-b 1,2
+-c 1,2
+--run-config-search-max-concurrency 2
+--run-config-search-max-instance-count 2
+--run-config-search-max-model-batch-size 2
\ No newline at end of file
diff --git a/src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-model-gpu.csv b/src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-model-gpu.csv
new file mode 100644
index 000000000..ce20b6e30
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-model-gpu.csv
@@ -0,0 +1,18 @@
+Model,GPU UUID,Batch,Concurrency,Model Config Path,Instance Group,Satisfies Constraints,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,2,add_sub_config_default,1:GPU,Yes,872.4,5.5,56.4
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,add_sub_config_default,1:GPU,Yes,872.4,4.2,56.3
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,1,add_sub_config_default,1:GPU,Yes,872.4,3.0,56.2
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,add_sub_config_default,1:GPU,Yes,872.4,2.0,56.2
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,2,add_sub_config_1,1:GPU,Yes,872.4,6.0,56.6
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,add_sub_config_1,1:GPU,Yes,872.4,4.8,56.6
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,1,add_sub_config_1,1:GPU,Yes,872.4,3.5,56.5
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,add_sub_config_1,1:GPU,Yes,872.4,1.8,56.4
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,2,add_sub_config_3,2:GPU,Yes,874.5,2.8,56.8
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,2,1,add_sub_config_3,2:GPU,Yes,874.5,3.0,56.8
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,add_sub_config_3,2:GPU,Yes,874.5,3.0,56.6
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,add_sub_config_3,2:GPU,Yes,874.5,1.7,56.7
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,add_sub_config_0,1:GPU,Yes,872.4,4.0,56.5
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,add_sub_config_0,1:GPU,Yes,872.4,2.0,56.4
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,2,add_sub_config_2,2:GPU,Yes,874.5,2.0,56.7
+add_sub,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,1,1,add_sub_config_2,2:GPU,Yes,874.5,3.0,56.6
+
diff --git a/src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-model-inference.csv b/src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-model-inference.csv
new file mode 100644
index 000000000..64d216513
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-model-inference.csv
@@ -0,0 +1,18 @@
+Model,Batch,Concurrency,Model Config Path,Instance Group,Max Batch Size,Satisfies Constraints,Throughput (infer/sec),p99 Latency (ms)
+add_sub,2,2,add_sub_config_default,1:GPU,8,Yes,13205.0,0.5
+add_sub,1,2,add_sub_config_default,1:GPU,8,Yes,6176.0,0.6
+add_sub,2,1,add_sub_config_default,1:GPU,8,Yes,4876.1,0.7
+add_sub,1,1,add_sub_config_default,1:GPU,8,Yes,2508.0,0.5
+add_sub,2,2,add_sub_config_1,1:GPU,2,Yes,12528.9,0.6
+add_sub,1,2,add_sub_config_1,1:GPU,2,Yes,6109.9,0.6
+add_sub,2,1,add_sub_config_1,1:GPU,2,Yes,4804.5,0.6
+add_sub,1,1,add_sub_config_1,1:GPU,2,Yes,2186.0,0.6
+add_sub,2,2,add_sub_config_3,2:GPU,2,Yes,6967.9,0.8
+add_sub,2,1,add_sub_config_3,2:GPU,2,Yes,3988.7,0.8
+add_sub,1,2,add_sub_config_3,2:GPU,2,Yes,3270.7,0.8
+add_sub,1,1,add_sub_config_3,2:GPU,2,Yes,1752.1,0.8
+add_sub,1,2,add_sub_config_0,1:GPU,1,Yes,6114.9,0.6
+add_sub,1,1,add_sub_config_0,1:GPU,1,Yes,2334.7,0.6
+add_sub,1,2,add_sub_config_2,2:GPU,1,Yes,3363.1,0.9
+add_sub,1,1,add_sub_config_2,2:GPU,1,Yes,1915.6,0.8
+
diff --git a/src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-server-only.csv b/src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-server-only.csv
new file mode 100644
index 000000000..f25e1df5c
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/single-model-ckpt/golden-metrics-server-only.csv
@@ -0,0 +1,3 @@
+Model,GPU UUID,GPU Memory Usage (MB),GPU Utilization (%),GPU Power Usage (W)
+triton-server,GPU-8557549f-9c89-4384-8bd6-1fd823c342e0,870.0,0.0,56.0
+
diff --git a/src/c++/perf_analyzer/tests/common/single-model-ckpt/plot_manager.json b/src/c++/perf_analyzer/tests/common/single-model-ckpt/plot_manager.json
new file mode 100644
index 000000000..cfd9c1a8b
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/single-model-ckpt/plot_manager.json
@@ -0,0 +1 @@
+{"_simple_plots": {"add_sub": {"throughput_v_latency": {"_name": "throughput_v_latency", "_title": "Throughput vs. Latency", "_x_axis": "perf_latency_p99", "_y_axis": "perf_throughput", "_monotonic": true, "_data": {"add_sub_config_default": {"x_data": [0.524, 0.581, 0.658, 0.539], "y_data": [13205.0, 6176.0, 4876.12, 2508.04]}, "add_sub_config_1": {"x_data": [0.596, 0.607, 0.587, 0.599], "y_data": [12528.9, 6109.95, 4804.46, 2185.97]}, "add_sub_config_3": {"x_data": [0.81, 0.803, 0.838, 0.839], "y_data": [6967.92, 3988.65, 3270.69, 1752.08]}}}, "gpu_mem_v_latency": {"_name": "gpu_mem_v_latency", "_title": "GPU Memory vs. Latency", "_x_axis": "perf_latency_p99", "_y_axis": "gpu_used_memory", "_monotonic": false, "_data": {"add_sub_config_default": {"x_data": [0.524, 0.581, 0.658, 0.539], "y_data": [872.415232, 872.415232, 872.415232, 872.415232]}, "add_sub_config_1": {"x_data": [0.596, 0.607, 0.587, 0.599], "y_data": [872.415232, 872.415232, 872.415232, 872.415232]}, "add_sub_config_3": {"x_data": [0.81, 0.803, 0.838, 0.839], "y_data": [874.512384, 874.512384, 874.512384, 874.512384]}}}}}}
\ No newline at end of file
diff --git a/src/c++/perf_analyzer/tests/common/test_result_collector.py b/src/c++/perf_analyzer/tests/common/test_result_collector.py
new file mode 100755
index 000000000..423481900
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/test_result_collector.py
@@ -0,0 +1,47 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import json
+import unittest
+
+
+class TestResultCollector(unittest.TestCase):
+    """
+    TestResultCollector stores test result and prints it to stdout. In order
+    to use this class, unit tests must inherit this class. Use
+    `check_test_results` bash function from `common/util.sh` to verify the
+    expected number of tests produced by this class
+    """
+
+    @classmethod
+    def setResult(cls, total, errors, failures):
+        cls.total, cls.errors, cls.failures = total, errors, failures
+
+    @classmethod
+    def tearDownClass(cls):
+        # this method is called when all the unit tests in a class are
+        # finished.
+        json_res = {"total": cls.total, "errors": cls.errors, "failures": cls.failures}
+        print(json.dumps(json_res))
+
+    def run(self, result=None):
+        # result argument stores the accumulative test results
+        test_result = super().run(result)
+        if test_result:
+            total = test_result.testsRun
+            errors = len(test_result.errors)
+            failures = len(test_result.failures)
+            self.setResult(total, errors, failures)
diff --git a/src/c++/perf_analyzer/tests/common/test_utils.py b/src/c++/perf_analyzer/tests/common/test_utils.py
new file mode 100755
index 000000000..9d418027f
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/common/test_utils.py
@@ -0,0 +1,525 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+from typing import Tuple, Union
+
+from model_analyzer.cli.cli import CLI
+from model_analyzer.config.input.config_command_profile import ConfigCommandProfile
+from model_analyzer.config.input.config_command_report import ConfigCommandReport
+from model_analyzer.config.input.config_defaults import (
+    DEFAULT_BATCH_SIZES,
+    DEFAULT_CLIENT_PROTOCOL,
+    DEFAULT_MEASUREMENT_MODE,
+    DEFAULT_MONITORING_INTERVAL,
+    DEFAULT_OUTPUT_MODEL_REPOSITORY,
+    DEFAULT_TRITON_GRPC_ENDPOINT,
+    DEFAULT_TRITON_HTTP_ENDPOINT,
+    DEFAULT_TRITON_INSTALL_PATH,
+    DEFAULT_TRITON_LAUNCH_MODE,
+    DEFAULT_TRITON_METRICS_URL,
+)
+from model_analyzer.config.run.model_run_config import ModelRunConfig
+from model_analyzer.config.run.run_config import RunConfig
+from model_analyzer.constants import SECONDS_TO_MILLISECONDS_MULTIPLIER
+from model_analyzer.perf_analyzer.perf_config import PerfAnalyzerConfig
+from model_analyzer.record.metrics_manager import MetricsManager
+from model_analyzer.result.constraint_manager import ConstraintManager
+from model_analyzer.result.result_manager import ResultManager
+from model_analyzer.result.run_config_measurement import RunConfigMeasurement
+from model_analyzer.result.run_config_result import RunConfigResult
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+from model_analyzer.triton.model.model_config import ModelConfig
+from model_analyzer.triton.model.model_config_variant import ModelConfigVariant
+from tests.mocks.mock_config import MockConfig
+
+ROOT_DIR = os.path.dirname(os.path.abspath(__file__))
+
+
+def evaluate_mock_config(
+    args: list, yaml_str: str, subcommand: str = "profile"
+) -> Union[ConfigCommandProfile, ConfigCommandReport]:
+    """
+    Return a ConfigCommandReport/Analyze/Profile created from the fake CLI
+    'args' list and fake config.yaml contents 'yaml_str'
+    """
+    yaml_content = convert_to_bytes(yaml_str)
+    mock_config = MockConfig(args, yaml_content)
+    mock_config.start()
+
+    if subcommand == "report":
+        config = ConfigCommandReport()
+    else:
+        config = ConfigCommandProfile()
+
+    cli = CLI()
+    cli.add_subcommand(cmd=subcommand, help="", config=config)
+    cli.parse()
+    mock_config.stop()
+    return config
+
+
+def load_single_model_result_manager() -> Tuple[ResultManager, ConfigCommandProfile]:
+    """
+    Return a ResultManager with the single model test checkpoint loaded, as well
+    as the ConfigCommandProfile used to fake the profile step
+    """
+    dir_path = f"{ROOT_DIR}/single-model-ckpt/"
+    yaml_str = "profile_models: add_sub"
+    return _load_result_manager_helper(dir_path, yaml_str)
+
+
+def load_multi_model_result_manager() -> Tuple[ResultManager, ConfigCommandProfile]:
+    """
+    Return a ResultManager with the multi model test checkpoint loaded, as well
+    as the ConfigCommandProfile used to fake the profile step
+    """
+    dir_path = f"{ROOT_DIR}/multi-model-ckpt/"
+    yaml_str = "profile_models: resnet50_libtorch,vgg19_libtorch"
+    return _load_result_manager_helper(dir_path, yaml_str)
+
+
+def load_ensemble_result_manager() -> Tuple[ResultManager, ConfigCommandProfile]:
+    """
+    Return a ResultManager with the ensemble model test checkpoint loaded, as well
+    as the ConfigCommandProfile used to fake the profile step
+    """
+    dir_path = f"{ROOT_DIR}/ensemble-ckpt/"
+    yaml_str = "profile_models: ensemble_add_sub"
+    return _load_result_manager_helper(dir_path, yaml_str)
+
+
+def load_bls_result_manager() -> Tuple[ResultManager, ConfigCommandProfile]:
+    """
+    Return a ResultManager with the BLS model test checkpoint loaded, as well
+    as the ConfigCommandProfile used to fake the profile step
+    """
+    dir_path = f"{ROOT_DIR}/bls-ckpt/"
+    yaml_str = "profile_models: bls"
+    return _load_result_manager_helper(dir_path, yaml_str)
+
+
+def load_request_rate_result_manager() -> Tuple[ResultManager, ConfigCommandProfile]:
+    """
+    Return a ResultManager with the request rate model test checkpoint loaded, as well
+    as the ConfigCommandProfile used to fake the profile step
+    """
+    dir_path = f"{ROOT_DIR}/request-rate-ckpt/"
+    yaml_str = f"""
+        request_rate_search_enable: true
+        profile_models: add_sub
+    """
+    return _load_result_manager_helper(dir_path, yaml_str)
+
+
+def _load_result_manager_helper(dir_path: str, yaml_str: str):
+    args = [
+        "model-analyzer",
+        "profile",
+        "-f",
+        "config.yml",
+        "--checkpoint-directory",
+        dir_path,
+        "--export-path",
+        dir_path,
+        "--model-repository",
+        ".",
+        "--run-config-profile-models-concurrently-enable",
+        "--run-config-search-mode",
+        "quick",
+    ]
+    config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+    state_manager = AnalyzerStateManager(config=config, server=None)
+    state_manager.load_checkpoint(checkpoint_required=True)
+
+    result_manager = ResultManager(
+        config=config,
+        state_manager=state_manager,
+        constraint_manager=ConstraintManager(config=config),
+    )
+    return result_manager, config
+
+
+def convert_to_bytes(string):
+    """
+    Converts string into bytes and ensures minimum length requirement
+    for compatibility with unpack function called in usr/lib/python3.8/gettext.py
+
+    Parameters
+    ----------
+    string: str
+    """
+    if len(string) > 4:
+        return bytes(string, "utf-8")
+    else:
+        return bytes(string + "    ", "utf-8")
+
+
+def convert_non_gpu_metrics_to_data(non_gpu_metric_values):
+    """
+    Non GPU data will be a dict whose keys and values are
+    a list of Records
+
+    Parameters
+    ----------
+    non_gpu_metric_values: dict of non-gpu metrics
+    """
+
+    non_gpu_data = []
+    non_gpu_metric_tags = list(non_gpu_metric_values.keys())
+
+    for i, metric in enumerate(MetricsManager.get_metric_types(non_gpu_metric_tags)):
+        non_gpu_data.append(metric(value=non_gpu_metric_values[non_gpu_metric_tags[i]]))
+
+    return non_gpu_data
+
+
+def convert_gpu_metrics_to_data(gpu_metric_values):
+    """
+    GPU data will be a dict whose keys are gpu_ids and values
+    are lists of Records
+
+    Parameters
+    ----------
+    gpu_metric_values: dict of gpu metrics
+    """
+    gpu_data = {}
+    for gpu_uuid, metrics_values in gpu_metric_values.items():
+        gpu_data[gpu_uuid] = []
+        gpu_metric_tags = list(metrics_values.keys())
+        for i, gpu_metric in enumerate(
+            MetricsManager.get_metric_types(gpu_metric_tags)
+        ):
+            gpu_data[gpu_uuid].append(
+                gpu_metric(value=metrics_values[gpu_metric_tags[i]])
+            )
+
+    return gpu_data
+
+
+def convert_avg_gpu_metrics_to_data(avg_gpu_metric_values):
+    """
+    Avg GPU data will be a dict of Records
+
+    Parameters
+    ----------
+    gpu_metric_values: dict of gpu metrics
+    """
+    avg_gpu_data = {}
+    avg_gpu_metric_tags = list(avg_gpu_metric_values.keys())
+
+    for i, avg_gpu_metric in enumerate(
+        MetricsManager.get_metric_types(avg_gpu_metric_tags)
+    ):
+        avg_gpu_data[avg_gpu_metric_tags[i]] = avg_gpu_metric(
+            value=avg_gpu_metric_values[avg_gpu_metric_tags[i]]
+        )
+
+    return avg_gpu_data
+
+
+def construct_perf_analyzer_config(
+    model_name="my-model",
+    output_file_name="my-model-results.csv",
+    batch_size=DEFAULT_BATCH_SIZES,
+    concurrency=1,
+    request_rate=None,
+    launch_mode=DEFAULT_TRITON_LAUNCH_MODE,
+    client_protocol=DEFAULT_CLIENT_PROTOCOL,
+    perf_analyzer_flags=None,
+):
+    """
+    Constructs a Perf Analyzer Config
+
+    Parameters
+    ----------
+    model_name: str
+        The name of the model
+    output_file_name: str
+        The name of the output file
+    batch_size: int
+        The batch size for this PA configuration
+    concurrency: int
+        The concurrency value for this PA configuration
+    request_rate: int
+        The request rate value for this PA configuration
+    launch_mode: str
+        The launch mode for this PA configuration
+    client_protocol: str
+        The client protocol for this PA configuration
+    perf_analyzer_flags: dict
+        A dict of any additional PA flags to be set
+
+    Returns
+    -------
+    PerfAnalyzerConfig
+        constructed with all of the above data.
+    """
+
+    pa_config = PerfAnalyzerConfig()
+    pa_config._options["-m"] = model_name
+    pa_config._options["-f"] = output_file_name
+    pa_config._options["-b"] = batch_size
+
+    if request_rate:
+        pa_config._args["request-rate-range"] = request_rate
+    else:
+        pa_config._args["concurrency-range"] = concurrency
+
+    pa_config._args["measurement-mode"] = DEFAULT_MEASUREMENT_MODE
+
+    pa_config.update_config(perf_analyzer_flags)
+
+    if launch_mode == "c_api":
+        pa_config._args["service-kind"] = "triton_c_api"
+        pa_config._args["triton-server-directory"] = DEFAULT_TRITON_INSTALL_PATH
+        pa_config._args["model-repository"] = DEFAULT_OUTPUT_MODEL_REPOSITORY
+    else:
+        pa_config._args["collect-metrics"] = "True"
+        pa_config._args["metrics-url"] = DEFAULT_TRITON_METRICS_URL
+        pa_config._args["metrics-interval"] = (
+            SECONDS_TO_MILLISECONDS_MULTIPLIER * DEFAULT_MONITORING_INTERVAL
+        )
+        pa_config._options["-i"] = client_protocol
+        if client_protocol == "http":
+            pa_config._options["-u"] = DEFAULT_TRITON_HTTP_ENDPOINT
+        else:
+            pa_config._options["-u"] = DEFAULT_TRITON_GRPC_ENDPOINT
+
+    return pa_config
+
+
+def construct_run_config(
+    model_name: str, model_config_variant_name: str, pa_config_name: str
+) -> RunConfig:
+    """
+    Constructs a Perf Analyzer Config
+    """
+
+    model_config_dict = {"name": model_name}
+    model_config = ModelConfig.create_from_dictionary(model_config_dict)
+    model_config_variant = ModelConfigVariant(model_config, model_config_variant_name)
+
+    perf_config = PerfAnalyzerConfig()
+    perf_config.update_config({"model-name": pa_config_name})
+    mrc = ModelRunConfig(model_name, model_config_variant, perf_config)
+    rc = RunConfig({})
+    rc.add_model_run_config(mrc)
+    return rc
+
+
+def construct_run_config_measurement(
+    model_name,
+    model_config_names,
+    model_specific_pa_params,
+    gpu_metric_values,
+    non_gpu_metric_values,
+    constraint_manager=None,
+    metric_objectives=None,
+    model_config_weights=None,
+):
+    """
+    Construct a RunConfig measurement from the given data
+
+    Parameters
+    ----------
+    model_name: str
+        The name of the model that generated this result
+    model_config_names: list of str
+        A list of Model Config names that generated this result
+    model_specific_pa_params: list of dict
+        A list (one per model config) of dict's of PA parameters that change
+        between models in a multi-model run
+    gpu_metric_values: dict
+        Keys are gpu id, values are dict
+        The dict where keys are gpu based metric tags, values are the data
+    non_gpu_metric_values: list of dict
+        List of (one per model config) dict's where keys are non gpu perf metrics, values are the data
+    metric_objectives: list of RecordTypes
+        A list of metric objectives (one per model config) used to compare measurements
+    model_config_weights: list of ints
+        A list of weights (one per model config) used to bias measurement results between models
+
+    Returns
+    -------
+    RunConfigMeasurement
+        constructed with all of the above data.
+    """
+
+    gpu_data = convert_gpu_metrics_to_data(gpu_metric_values)
+
+    model_variants_name = "".join(model_config_names)
+    rc_measurement = RunConfigMeasurement(model_variants_name, gpu_data)
+
+    non_gpu_data = [
+        convert_non_gpu_metrics_to_data(non_gpu_metric_value)
+        for non_gpu_metric_value in non_gpu_metric_values
+    ]
+
+    for index, model_config_name in enumerate(model_config_names):
+        rc_measurement.add_model_config_measurement(
+            model_config_name=model_config_name,
+            model_specific_pa_params=model_specific_pa_params[index],
+            non_gpu_data=non_gpu_data[index],
+        )
+
+    if model_config_weights:
+        rc_measurement.set_model_config_weighting(model_config_weights)
+
+    if metric_objectives:
+        rc_measurement.set_metric_weightings(metric_objectives)
+
+    if constraint_manager:
+        rc_measurement.set_constraint_manager(constraint_manager=constraint_manager)
+
+    return rc_measurement
+
+
+def construct_run_config_result(
+    avg_gpu_metric_values,
+    avg_non_gpu_metric_values_list,
+    comparator,
+    constraint_manager=None,
+    value_step=1,
+    model_name="test_model",
+    model_config_names=["test_model"],
+    run_config=None,
+):
+    """
+    Takes a dictionary whose values are average
+    metric values, constructs artificial data
+    around these averages, and then constructs
+    a result from this data.
+
+    Parameters
+    ----------
+    avg_gpu_metric_values: dict
+        The dict where keys are gpu based metric tags
+        and values are the average values around which
+        we want data
+    avg_non_gpu_metric_values: list of dict
+        Per model list of:
+            keys are non gpu perf metrics, values are their
+            average values.
+    value_step: int
+        The step value between two adjacent data values.
+        Can be used to control the max/min of the data
+        distribution in the construction result
+    comparator: RunConfigResultComparator
+        The comparator used to compare measurements/results
+    model_name: str
+        The name of the model that generated this result
+    model_config: ModelConfig
+        The model config used to generate this result
+    """
+
+    num_vals = 10
+
+    # Construct a result
+    run_config_result = RunConfigResult(
+        model_name=model_name,
+        run_config=run_config,
+        comparator=comparator,
+        constraint_manager=constraint_manager,
+    )
+
+    # Get dict of list of metric values
+    gpu_metric_values = {}
+    for gpu_uuid, metric_values in avg_gpu_metric_values.items():
+        gpu_metric_values[gpu_uuid] = {
+            key: list(
+                range(
+                    val - value_step * num_vals, val + value_step * num_vals, value_step
+                )
+            )
+            for key, val in metric_values.items()
+        }
+
+    non_gpu_metric_values_list = []
+    for avg_non_gpu_metric_values in avg_non_gpu_metric_values_list:
+        non_gpu_metric_values_list.append(
+            {
+                key: list(
+                    range(
+                        val - value_step * num_vals,
+                        val + value_step * num_vals,
+                        value_step,
+                    )
+                )
+                for key, val in avg_non_gpu_metric_values.items()
+            }
+        )
+
+    # Construct measurements and add them to the result
+    for i in range(2 * num_vals):
+        gpu_metrics = {}
+        for gpu_uuid, metric_values in gpu_metric_values.items():
+            gpu_metrics[gpu_uuid] = {
+                key: metric_values[key][i] for key in metric_values
+            }
+
+        non_gpu_metrics = []
+        for non_gpu_metric_values in non_gpu_metric_values_list:
+            non_gpu_metrics.append(
+                {key: non_gpu_metric_values[key][i] for key in non_gpu_metric_values}
+            )
+
+        run_config_result.add_run_config_measurement(
+            construct_run_config_measurement(
+                model_name=model_name,
+                model_config_names=model_config_names,
+                model_specific_pa_params=[
+                    {"batch_size": 1, "concurrency": 1}
+                    for model_config_name in model_config_names
+                ],
+                gpu_metric_values=gpu_metrics,
+                non_gpu_metric_values=non_gpu_metrics,
+                metric_objectives=comparator._metric_weights,
+            )
+        )
+
+    return run_config_result
+
+
+def construct_constraint_manager(yaml_config_str=None):
+    """
+    Returns a ConstraintManager object for Test cases
+
+    Parameters
+    ----------
+    yaml_config_str: optional str
+        yaml config string with at least one profile model name and optional constraints
+    """
+    args = ["model-analyzer", "profile", "-f", "config.yml", "-m", "."]
+
+    if not yaml_config_str:
+        yaml_config_str = """
+            profile_models:
+              test_model
+        """
+
+    config = evaluate_mock_config(args, yaml_config_str, subcommand="profile")
+
+    return ConstraintManager(config)
+
+
+def default_encode(obj):
+    if isinstance(obj, bytes):
+        return obj.decode("utf-8")
+    elif hasattr(obj, "to_dict"):
+        return obj.to_dict()
+    else:
+        return obj.__dict__
diff --git a/src/c++/perf_analyzer/tests/mocks/__init__.py b/src/c++/perf_analyzer/tests/mocks/__init__.py
new file mode 100755
index 000000000..f7f71eb3d
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/__init__.py
@@ -0,0 +1,15 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_api_error.py b/src/c++/perf_analyzer/tests/mocks/mock_api_error.py
new file mode 100755
index 000000000..a93d8c61a
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_api_error.py
@@ -0,0 +1,24 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+class MockAPIError(Exception):
+    """
+    Mock for Docker API errors
+    """
+
+    def __init__(self):
+        self.explanation = "port is already allocated"
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_base.py b/src/c++/perf_analyzer/tests/mocks/mock_base.py
new file mode 100755
index 000000000..0325220d7
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_base.py
@@ -0,0 +1,38 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from abc import ABC, abstractmethod
+
+
+class MockBase(ABC):
+    """
+    Base abstract class for all the mocks
+    """
+
+    def __init__(self):
+        self._patchers = []
+
+    @abstractmethod
+    def _fill_patchers(self):
+        pass
+
+    def start(self):
+        for patch in self._patchers:
+            patch.start()
+
+    def stop(self):
+        for patch in self._patchers:
+            patch.stop()
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_client.py b/src/c++/perf_analyzer/tests/mocks/mock_client.py
new file mode 100755
index 000000000..caaa191eb
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_client.py
@@ -0,0 +1,221 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import ANY, MagicMock, Mock, patch
+
+from tritonclient.http import InferenceServerException
+
+from .mock_base import MockBase
+
+
+class MockTritonClientMethods(MockBase):
+    """
+    Mocks the tritonclient module functions
+    used in model_analyzer/triton/client
+    Provides functions to check operation.
+    """
+
+    TEST_MODEL_CONFIG = {}
+
+    def __init__(self):
+        client_attrs = {
+            "load_model": MagicMock(),
+            "unload_model": MagicMock(),
+            "is_model_ready": MagicMock(return_value=True),
+            "is_server_ready": MagicMock(return_value=True),
+            "get_model_config": MagicMock(),
+        }
+        mock_http_client = Mock(**client_attrs)
+        mock_grpc_client = Mock(**client_attrs)
+        self.patcher_http_client = patch(
+            "model_analyzer.triton.client.http_client.httpclient.InferenceServerClient",
+            Mock(return_value=mock_http_client),
+        )
+        self.patcher_grpc_client = patch(
+            "model_analyzer.triton.client.grpc_client.grpcclient.InferenceServerClient",
+            Mock(return_value=mock_grpc_client),
+        )
+        super().__init__()
+        self._fill_patchers()
+
+    def start(self):
+        """
+        start the patchers
+        """
+
+        self.http_mock = self.patcher_http_client.start()
+        self.grpc_mock = self.patcher_grpc_client.start()
+
+    def _fill_patchers(self):
+        """
+        Fills the patcher list for destruction
+        """
+
+        self._patchers.append(self.patcher_http_client)
+        self._patchers.append(self.patcher_grpc_client)
+
+    def assert_created_grpc_client_with_args(
+        self,
+        url,
+        ssl=False,
+        root_certificates=None,
+        private_key=None,
+        certificate_chain=None,
+    ):
+        """
+        Assert that the correct InferServerClient was
+        indeed constructed with the specified url and SSL options
+        """
+
+        self.grpc_mock.assert_called_with(
+            url=url,
+            ssl=ssl,
+            root_certificates=root_certificates,
+            private_key=private_key,
+            certificate_chain=certificate_chain,
+        )
+
+    def assert_created_http_client_with_args(
+        self, url, ssl_options={}, ssl=False, ssl_context_factory=ANY, insecure=True
+    ):
+        """
+        Assert that the correct InferServerClient was
+        indeed constructed with the specified url  and SSL options
+        """
+
+        self.http_mock.assert_called_with(
+            url=url,
+            ssl_options=ssl_options,
+            ssl=ssl,
+            ssl_context_factory=ssl_context_factory,
+            insecure=insecure,
+        )
+
+    def assert_grpc_client_waited_for_server_ready(self):
+        """
+        Assert that the correct InferServerClient
+        indeed called is_server_ready
+        """
+
+        self.grpc_mock.return_value.is_server_ready.assert_called()
+
+    def assert_http_client_waited_for_server_ready(self):
+        """
+        Assert that the correct InferServerClient
+        indeed called is_server_ready
+        """
+
+        self.http_mock.return_value.is_server_ready.assert_called()
+
+    def assert_grpc_client_waited_for_model_ready(self, model_name):
+        """
+        Assert that the correct InferServerClient
+        indeed called is_model_ready with correct model
+        """
+
+        self.grpc_mock.return_value.is_model_ready.assert_called_with(model_name)
+
+    def assert_http_client_waited_for_model_ready(self, model_name):
+        """
+        Assert that the correct InferServerClient
+        indeed called is_model_ready with correct model
+        """
+
+        self.http_mock.return_value.is_model_ready.assert_called_with(model_name)
+
+    def raise_exception_on_load(self):
+        """
+        Set load_model to throw
+        InferenceServerException
+        """
+
+        self.grpc_mock.return_value.load_model.side_effect = InferenceServerException(
+            ""
+        )
+        self.http_mock.return_value.load_model.side_effect = InferenceServerException(
+            ""
+        )
+
+    def raise_exception_on_unload(self):
+        """
+        Set unload_model to throw
+        InferenceServerException
+        """
+
+        self.grpc_mock.return_value.unload_model.side_effect = Exception
+        self.http_mock.return_value.unload_model.side_effect = Exception
+
+    def raise_exception_on_wait_for_server_ready(self):
+        """
+        Set is_server_ready to throw
+        InferenceServerException
+        """
+
+        self.grpc_mock.return_value.is_server_ready.side_effect = Exception
+        self.http_mock.return_value.is_server_ready.side_effect = Exception
+
+    def raise_exception_on_wait_for_model_ready(self):
+        """
+        Set is_model_ready to throw
+        InferenceServerException
+        """
+
+        self.grpc_mock.return_value.is_model_ready.side_effect = Exception
+        self.http_mock.return_value.is_model_ready.side_effect = Exception
+
+    def set_server_not_ready(self):
+        """
+        Set is_server_ready's return value to False
+        """
+
+        self.grpc_mock.return_value.is_server_ready.return_value = False
+        self.http_mock.return_value.is_server_ready.return_value = False
+
+    def set_model_not_ready(self):
+        """
+        Set is_model_ready's return value to False
+        """
+
+        self.grpc_mock.return_value.is_model_ready.return_value = False
+        self.http_mock.return_value.is_model_ready.return_value = False
+
+    def set_model_config(self, model_config_dict):
+        """
+        Sets the value returned by client.get_model_config
+        """
+
+        self.grpc_mock.return_value.get_model_config.return_value = model_config_dict
+        self.http_mock.return_value.get_model_config.return_value = model_config_dict
+
+    def reset(self):
+        """
+        Reset the client mocks
+        """
+
+        self.grpc_mock.return_value.load_model.side_effect = None
+        self.http_mock.return_value.load_model.side_effect = None
+        self.grpc_mock.return_value.unload_model.side_effect = None
+        self.http_mock.return_value.unload_model.side_effect = None
+        self.grpc_mock.return_value.is_server_ready.side_effect = None
+        self.http_mock.return_value.is_server_ready.side_effect = None
+        self.grpc_mock.return_value.is_model_ready.side_effect = None
+        self.http_mock.return_value.is_model_ready.side_effect = None
+        self.grpc_mock.return_value.is_server_ready.return_value = True
+        self.http_mock.return_value.is_server_ready.return_value = True
+        self.grpc_mock.return_value.is_model_ready.return_value = True
+        self.http_mock.return_value.is_model_ready.return_value = True
+        self.grpc_mock.return_value.get_model_config.return_value = None
+        self.http_mock.return_value.get_model_config.return_value = None
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_config.py b/src/c++/perf_analyzer/tests/mocks/mock_config.py
new file mode 100755
index 000000000..cc8f1e4fb
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_config.py
@@ -0,0 +1,36 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, mock_open, patch
+
+from .mock_base import MockBase
+
+
+class MockConfig(MockBase):
+    def __init__(self, args, yaml_file_content):
+        self.args = args
+        self.yaml_file_content = yaml_file_content
+        super().__init__()
+        self._fill_patchers()
+
+    def _fill_patchers(self):
+        patchers = self._patchers
+
+        patchers.append(
+            patch("builtins.open", mock_open(read_data=self.yaml_file_content))
+        )
+        patchers.append(patch("sys.argv", self.args))
+        patchers.append(patch("numba.cuda.is_available", MagicMock(True)))
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_dcgm.py b/src/c++/perf_analyzer/tests/mocks/mock_dcgm.py
new file mode 100755
index 000000000..37bcf3ce6
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_dcgm.py
@@ -0,0 +1,55 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, patch
+
+from .mock_base import MockBase
+from .mock_dcgm_agent import MockDCGMAgent
+from .mock_dcgm_field_group_watcher import MockDCGMFieldGroupWatcherHelper
+
+
+class MockDCGM(MockBase):
+    """
+    Mocks dcgm_agent methods.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self._fill_patchers()
+
+    def _fill_patchers(self):
+        patchers = self._patchers
+
+        structs_imports_path = [
+            "model_analyzer.monitor.dcgm.dcgm_monitor",
+            "model_analyzer.device.gpu_device_factory",
+        ]
+        for import_path in structs_imports_path:
+            patchers.append(patch(f"{import_path}.structs._dcgmInit", MagicMock()))
+
+        dcgm_agent_imports_path = [
+            "model_analyzer.monitor.dcgm.dcgm_monitor",
+            "model_analyzer.device.gpu_device_factory",
+        ]
+        for import_path in dcgm_agent_imports_path:
+            patchers.append(patch(f"{import_path}.dcgm_agent", MockDCGMAgent))
+
+        patchers.append(
+            patch(
+                "model_analyzer.monitor.dcgm.dcgm_monitor.dcgm_field_helpers.DcgmFieldGroupWatcher",
+                MockDCGMFieldGroupWatcherHelper,
+            )
+        )
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_dcgm_agent.py b/src/c++/perf_analyzer/tests/mocks/mock_dcgm_agent.py
new file mode 100755
index 000000000..92faa2f7e
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_dcgm_agent.py
@@ -0,0 +1,197 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock
+
+import model_analyzer.monitor.dcgm.dcgm_structs as structs
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+
+TEST_PCI_BUS_ID = "00000000:6A:00.0"
+TEST_UUID = "dddddddd-bbbb-aaaa-cccc-ffffffffffff"
+
+
+class MockDCGMAgent:
+    device_groups = {}
+    field_groups = {}
+    devices = [{"pci_bus_id": TEST_PCI_BUS_ID, "uuid": TEST_UUID}]
+
+    @staticmethod
+    def dcgmInit():
+        """
+        Mock dcgmInit method
+        """
+
+        pass
+
+    @staticmethod
+    def dcgmStartEmbedded(op_mode):
+        """
+        Mock dcgmStartEmbedded
+
+        Parameters
+        ----------
+        op_mode : int
+            DCGM operational mode
+        """
+
+        pass
+
+    @staticmethod
+    def dcgmGetAllSupportedDevices(dcgm_handle):
+        """
+        Mock dcgmGetAllSupportedDevices
+
+        Parameters
+        ----------
+        dcgm_handle : ctypes.c_void_p
+            A DCGM Handle object
+        """
+
+        return list(range(0, len(MockDCGMAgent.devices)))
+
+    @staticmethod
+    def dcgmGetDeviceAttributes(dcgm_handle, device):
+        """
+        Mock dcgmGetDeviceAttributes
+
+        Parameters
+        ----------
+        dcgm_handle : ctypes.c_void_p
+            A DCGM Handle object
+        device : int
+            device id
+
+        Returns
+        -------
+        MagicMock
+            A MagicMock containing the device attributes
+        """
+
+        gpu_device = MockDCGMAgent.devices[device]
+        device_attribute = MagicMock()
+        device_attribute.identifiers.pciBusId = gpu_device["pci_bus_id"]
+        device_attribute.identifiers.uuid = gpu_device["uuid"]
+        return device_attribute
+
+    @staticmethod
+    def dcgmGroupCreate(dcgm_handle, type_name, name):
+        """
+        Mock dcgmGroupCreate
+
+        Parameters
+        ----------
+        dcgm_handle : ctypes.c_void_p
+            A DCGM Handle object
+        type_name : int
+            Group type
+        name : str
+            Device group name
+
+        Returns
+        -------
+        int
+            Returns the group id
+
+        Raises
+        ------
+        KeyError
+            If the group already exists it raises a KeyError
+        """
+
+        if not (name in MockDCGMAgent.device_groups):
+            if type_name == structs.DCGM_GROUP_EMPTY:
+                MockDCGMAgent.device_groups[name] = []
+                group_id = MagicMock()
+                group_id.value = list(MockDCGMAgent.device_groups).index(name)
+                return group_id
+        else:
+            raise KeyError
+
+    @staticmethod
+    def dcgmGroupAddDevice(dcgm_handle, group_id, gpu_device_id):
+        """
+        Mock dcgmGroupAddDevice
+
+        Parameters
+        ----------
+        dcgm_handle : ctypes.c_void_p
+            A DCGM Handle object
+        group_id : int
+            Group type
+        gpu_device_id : int
+            GPU device id
+
+        Raises
+        ------
+        KeyError
+            If the group does not exist
+        """
+
+        group_id = group_id.value
+        if group_id >= len(list(MockDCGMAgent.device_groups)):
+            raise KeyError
+
+        device_group_name = list(MockDCGMAgent.device_groups)[group_id]
+        device_group = MockDCGMAgent.device_groups[device_group_name]
+
+        if gpu_device_id in device_group:
+            raise TritonModelAnalyzerException(
+                f"GPU device {gpu_device_id} already exists in the device group"
+            )
+
+        device_group.append(gpu_device_id)
+
+    @staticmethod
+    def dcgmFieldGroupCreate(dcgm_handle, fields, name):
+        """
+        Mock dcgmFieldGroupCreate
+
+        Parameters
+        ----------
+        dcgm_handle : ctypes.c_void_p
+            A DCGM Handle object
+        fields : list
+            List of ints containing the fields to be monitored
+        name : str
+            Group name
+
+        Returns
+        -------
+        int
+            Returns the group id
+
+        Raises
+        ------
+        KeyError
+            If the group already exists it raises a KeyError
+        """
+
+        if not (name in MockDCGMAgent.field_groups):
+            MockDCGMAgent.field_groups[name] = fields
+            group_id = MagicMock()
+            group_id.value = list(MockDCGMAgent.field_groups).index(name)
+            return group_id
+        else:
+            raise KeyError
+
+    @staticmethod
+    def dcgmShutdown():
+        """
+        Mock dcgmShutdown
+        """
+
+        MockDCGMAgent.device_groups = {}
+        MockDCGMAgent.field_groups = {}
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_dcgm_field_group_watcher.py b/src/c++/perf_analyzer/tests/mocks/mock_dcgm_field_group_watcher.py
new file mode 100755
index 000000000..ed31705f8
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_dcgm_field_group_watcher.py
@@ -0,0 +1,94 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import time
+from collections import defaultdict
+from unittest.mock import MagicMock
+
+from .mock_dcgm_agent import MockDCGMAgent
+
+TEST_RECORD_VALUE = 2.4
+
+
+class MockDCGMFieldGroupWatcherHelper:
+    """
+    Mock of the DCGMFieldGroupWatcher class
+    """
+
+    def __init__(
+        self,
+        handle,
+        group_id,
+        field_group,
+        operation_mode,
+        update_freq,
+        max_keep_age,
+        max_keep_samples,
+        start_timestamp,
+    ):
+        """
+        handle : dcgm_handle
+            DCGM handle from dcgm_agent.dcgmInit()
+        groupId : int
+            a DCGM group ID returned from dcgm_agent.dcgmGroupCreate
+        fieldGroup : int
+            DcgmFieldGroup() instance to watch fields for
+        operationMode : dcgm_structs.DCGM_OPERATION_MODE
+            a dcgm_structs.DCGM_OPERATION_MODE_? constant for if the host
+            engine is running in lock step or auto mode
+        updateFreq : float
+            how often to update each field in usec
+        maxKeepAge : int
+            how long DCGM should keep values for in seconds
+        maxKeepSamples : int
+            is the maximum number of samples DCGM should ever cache for each
+            field
+        startTimestamp : int
+            a base timestamp we should start from when first reading
+            values. This can be used to resume a previous instance of a
+            DcgmFieldGroupWatcher by using its _nextSinceTimestamp. 0=start
+            with all cached data
+        """
+
+        self._handle = handle
+        self._group_id = group_id.value
+        self._field_group = field_group
+        self._operation_mode = operation_mode
+        self._update_freq = update_freq
+        self._max_keep_age = max_keep_age
+        self._max_keep_samples = max_keep_samples
+        self._start_timestamp = start_timestamp
+        self.values = defaultdict(lambda: defaultdict(MagicMock))
+
+    def GetMore(self):
+        """
+        This function performs a single iteration of monitoring
+        """
+
+        group_name = list(MockDCGMAgent.device_groups)[self._group_id]
+        device_group = MockDCGMAgent.device_groups[group_name]
+        field_group_name = list(MockDCGMAgent.field_groups)[self._field_group]
+
+        for device in device_group:
+            for field in MockDCGMAgent.field_groups[field_group_name]:
+                # Sample Record
+                record = MagicMock()
+                record.value = TEST_RECORD_VALUE
+                record.ts = int(time.time() * 1e6)
+                if not isinstance(self.values[device][field].values, list):
+                    self.values[device][field].values = [record]
+                else:
+                    self.values[device][field].values.append(record)
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_glob.py b/src/c++/perf_analyzer/tests/mocks/mock_glob.py
new file mode 100755
index 000000000..ccc51a549
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_glob.py
@@ -0,0 +1,54 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, Mock, patch
+
+from .mock_base import MockBase
+
+
+class MockGlobMethods(MockBase):
+    """
+    Mocks the methods for the os module
+    """
+
+    def __init__(self):
+        glob_attrs = {"glob": MagicMock(return_value=[])}
+        self.patcher_glob = patch(
+            "model_analyzer.state.analyzer_state_manager.glob", Mock(**glob_attrs)
+        )
+        super().__init__()
+        self._fill_patchers()
+
+    def start(self):
+        """
+        start the patchers
+        """
+
+        self.glob_mock = self.patcher_glob.start()
+
+    def _fill_patchers(self):
+        """
+        Fills the patcher list for destruction
+        """
+
+        self._patchers.append(self.patcher_glob)
+
+    def set_glob_return_value(self, value):
+        """
+        Sets the return value for the glob.glob call
+        """
+
+        self.glob_mock.glob.return_value = value
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_gpu_device_factory.py b/src/c++/perf_analyzer/tests/mocks/mock_gpu_device_factory.py
new file mode 100755
index 000000000..464d07d56
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_gpu_device_factory.py
@@ -0,0 +1,34 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, patch
+
+from .mock_base import MockBase
+
+
+class MockGPUDeviceFactory(MockBase):
+    def __init__(self, gpus):
+        self._cuda_visible_gpus = gpus
+        super().__init__()
+        self._fill_patchers()
+
+    def _fill_patchers(self):
+        self._patchers.append(
+            patch(
+                "model_analyzer.device.gpu_device_factory.GPUDeviceFactory.get_cuda_visible_gpus",
+                MagicMock(return_value=dict.fromkeys(self._cuda_visible_gpus)),
+            )
+        )
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_io.py b/src/c++/perf_analyzer/tests/mocks/mock_io.py
new file mode 100755
index 000000000..74a9bdf2c
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_io.py
@@ -0,0 +1,105 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, mock_open, patch
+
+from .mock_base import MockBase
+
+
+class MockIOMethods(MockBase):
+    """
+    A class that mocks filesystem
+    operations open and write
+    """
+
+    def __init__(self, mock_paths, read_data=[]):
+        self._mock_paths = mock_paths
+        self._patchers_open = {}
+        self._patchers_print = {}
+        self._open_mocks = {}
+        self._print_mocks = {}
+        for i in range(len(mock_paths)):
+            path = mock_paths[i]
+            data = read_data[i] if i < len(read_data) else None
+            self._patchers_open[path] = patch(f"{path}.open", mock_open(read_data=data))
+            self._patchers_print[path] = patch(f"{path}.print", MagicMock())
+        super().__init__()
+        self._fill_patchers()
+
+    def start(self):
+        """
+        start the patchers
+        """
+
+        for path in self._patchers_open:
+            self._open_mocks[path] = self._patchers_open[path].start()
+            self._print_mocks[path] = self._patchers_print[path].start()
+
+    def _fill_patchers(self):
+        """
+        Loads patchers into list
+        """
+
+        for patcher in self._patchers_open.values():
+            self._patchers.append(patcher)
+
+        for patcher in self._patchers_print.values():
+            self._patchers.append(patcher)
+
+    def raise_exception_on_open(self):
+        """
+        Raises an OSError when open is called
+        """
+
+        for mock in self._open_mocks.values():
+            mock.side_effect = OSError
+
+    def raise_exception_on_write(self):
+        """
+        Raises an OSError when write is called
+        """
+
+        for mock in self._open_mocks.values():
+            mock.return_value.write.side_effect = OSError
+
+    def assert_open_called_with_args(self, path, filename):
+        """
+        Asserts that file open was called
+        with given arguments
+        """
+
+        self._open_mocks[path].assert_called_with(filename)
+
+    def assert_write_called_with_args(self, path, out):
+        """
+        Asserts that file write was called
+        with given arguments
+        """
+
+        self._open_mocks[path].return_value.write.assert_called_with(out)
+
+    def assert_print_called_with_args(self, path, out):
+        """
+        Asserts that print was called
+        with given arguments
+        """
+
+        self._print_mocks[path].assert_called_with(out, end="")
+
+    def reset(self):
+        for mock in self._open_mocks.values():
+            mock.side_effect = None
+            mock.return_value.write.side_effect = None
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_json.py b/src/c++/perf_analyzer/tests/mocks/mock_json.py
new file mode 100755
index 000000000..968e19b6e
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_json.py
@@ -0,0 +1,61 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, Mock, patch
+
+from .mock_base import MockBase
+
+
+class MockJSONMethods(MockBase):
+    """
+    Mocks the methods for the os module
+    """
+
+    def __init__(self):
+        json_attrs = {"load": MagicMock(), "dump": MagicMock()}
+        self.patcher_json = patch(
+            "model_analyzer.state.analyzer_state_manager.json", Mock(**json_attrs)
+        )
+        super().__init__()
+        self._fill_patchers()
+
+    def start(self):
+        """
+        start the patchers
+        """
+
+        self.json_mock = self.patcher_json.start()
+
+    def _fill_patchers(self):
+        """
+        Fills the patcher list for destruction
+        """
+
+        self._patchers.append(self.patcher_json)
+
+    def set_json_load_return_value(self, value):
+        """
+        Sets the return value for json load
+        """
+
+        self.json_mock.load.return_value = value
+
+    def set_json_load_side_effect(self, effect):
+        """
+        Sets a side effet
+        """
+
+        self.json_mock.load.side_effect = effect
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_matplotlib.py b/src/c++/perf_analyzer/tests/mocks/mock_matplotlib.py
new file mode 100755
index 000000000..0ba8da3bc
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_matplotlib.py
@@ -0,0 +1,71 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, Mock, patch
+
+from .mock_base import MockBase
+
+
+class MockMatplotlibMethods(MockBase):
+    """
+    Mock class that mocks
+    matplotlib
+    """
+
+    def __init__(self):
+        plt_attrs = {"subplots": Mock(return_value=(MagicMock(), MagicMock()))}
+        self.patcher_pyplot = patch(
+            "model_analyzer.plots.simple_plot.plt", Mock(**plt_attrs)
+        )
+        super().__init__()
+        self._fill_patchers()
+
+    def start(self):
+        """
+        Start mock
+        """
+
+        self.pyplot_mock = self.patcher_pyplot.start()
+
+    def _fill_patchers(self):
+        """
+        Add patchers to list
+        """
+
+        self._patchers.append(self.patcher_pyplot)
+
+    def assert_called_subplots(self):
+        """
+        Checks for a call to subplots
+        """
+
+        self.pyplot_mock.subplots.assert_called()
+
+    def assert_called_plot_with_args(self, x_data, y_data, marker, label):
+        """
+        Checks for call to axes.plot
+        """
+
+        self.pyplot_mock.subplots.return_value[1].plot.assert_called_with(
+            x_data, y_data, marker=marker, label=label
+        )
+
+    def assert_called_save_with_args(self, filepath):
+        """
+        Checks for call to figure.savefig
+        """
+
+        self.pyplot_mock.subplots.return_value[0].savefig.assert_called_with(filepath)
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_model_config.py b/src/c++/perf_analyzer/tests/mocks/mock_model_config.py
new file mode 100755
index 000000000..4597dadba
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_model_config.py
@@ -0,0 +1,49 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, mock_open, patch
+
+from .mock_base import MockBase
+
+
+class MockModelConfig(MockBase):
+    def __init__(self, model_file_content=None):
+        self._model_file_content = model_file_content
+        super().__init__()
+        self._fill_patchers()
+
+    def _fill_patchers(self):
+        patchers = self._patchers
+
+        patchers.append(
+            patch("builtins.open", mock_open(read_data=self._model_file_content))
+        )
+        patchers.append(
+            patch(
+                "model_analyzer.triton.model.model_config.os.path.exists",
+                MagicMock(return_value=True),
+            )
+        )
+
+        def isfile(file_name):
+            if file_name.endswith(".pbtxt"):
+                return True
+            else:
+                return False
+
+        patchers.append(
+            patch("model_analyzer.triton.model.model_config.os.path.isfile", isfile)
+        )
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_numba.py b/src/c++/perf_analyzer/tests/mocks/mock_numba.py
new file mode 100755
index 000000000..070312cdc
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_numba.py
@@ -0,0 +1,66 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, Mock, patch
+
+from .mock_base import MockBase
+from .mock_dcgm_agent import TEST_PCI_BUS_ID
+
+
+class MockNumba(MockBase):
+    """
+    Mocks numba class
+    """
+
+    def __init__(self, mock_paths, is_available=True):
+        device = MagicMock()
+
+        # Ignore everything after 0
+        test_pci_id = TEST_PCI_BUS_ID.split(".")[0]
+
+        pci_domain_id, pci_bus_id, pci_device_id = test_pci_id.split(":")
+        device.get_device_identity = MagicMock(
+            return_value={
+                "pci_bus_id": int(pci_bus_id, 16),
+                "pci_domain_id": int(pci_domain_id, 16),
+                "pci_device_id": int(pci_device_id, 16),
+            }
+        )
+        device.id = 0
+        list_devices = MagicMock()
+        list_devices.return_value = [device]
+
+        cuda_attrs = {
+            "list_devices": list_devices,
+            "is_available": MagicMock(return_value=is_available),
+        }
+        numba_attrs = {"cuda": MagicMock(**cuda_attrs)}
+        self._mock_paths = mock_paths
+        self._patchers_numba = {}
+        self._numba_mocks = {}
+
+        for path in mock_paths:
+            self._patchers_numba[path] = patch(f"{path}.numba", Mock(**numba_attrs))
+        super().__init__()
+        self._fill_patchers()
+
+    def _fill_patchers(self):
+        """
+        Fills the patcher list for destruction
+        """
+
+        for patcher in self._patchers_numba.values():
+            self._patchers.append(patcher)
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_os.py b/src/c++/perf_analyzer/tests/mocks/mock_os.py
new file mode 100755
index 000000000..3187a20c9
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_os.py
@@ -0,0 +1,91 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+from unittest.mock import MagicMock, Mock, patch
+
+from .mock_base import MockBase
+
+
+class MockOSMethods(MockBase):
+    """
+    Mocks the methods for the os module
+    """
+
+    def __init__(self, mock_paths):
+        path_attrs = {
+            "join": MagicMock(return_value=""),
+            "abspath": MagicMock(return_value=""),
+            "isdir": MagicMock(return_value=True),
+            "exists": MagicMock(return_value=True),
+            "isfile": MagicMock(return_value=True),
+            "split": os.path.split,
+        }
+        os_attrs = {
+            "path": Mock(**path_attrs),
+            "mkdir": MagicMock(),
+            "makedirs": MagicMock(),
+            "getenv": MagicMock(),
+            "listdir": MagicMock(return_value=[]),
+            "environ": {"VARIABLE": "value"},
+        }
+        self._mock_paths = mock_paths
+        self._patchers_os = {}
+        self._os_mocks = {}
+        for path in mock_paths:
+            self._patchers_os[path] = patch(f"{path}.os", Mock(**os_attrs))
+        super().__init__()
+        self._fill_patchers()
+
+    def start(self):
+        """
+        start the patchers
+        """
+
+        for path in self._mock_paths:
+            self._os_mocks[path] = self._patchers_os[path].start()
+
+    def _fill_patchers(self):
+        """
+        Fills the patcher list for destruction
+        """
+
+        for patcher in self._patchers_os.values():
+            self._patchers.append(patcher)
+
+    def set_os_path_exists_return_value(self, value):
+        """
+        Sets the return value for os.path.exists
+        """
+
+        for mock in self._os_mocks.values():
+            mock.path.exists.return_value = value
+
+    def set_os_path_join_return_value(self, value):
+        """
+        Sets the return value for os.path.join
+        """
+
+        for mock in self._os_mocks.values():
+            mock.path.join.return_value = value
+
+    def set_os_path_isfile_return_value(self, value):
+        """
+        Sets the return value for os.path.isfile
+        """
+
+        for mock in self._os_mocks.values():
+            mock.path.isfile.return_value = value
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_perf_analyzer.py b/src/c++/perf_analyzer/tests/mocks/mock_perf_analyzer.py
new file mode 100755
index 000000000..4b656fd1e
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_perf_analyzer.py
@@ -0,0 +1,104 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, Mock, patch
+
+from .mock_base import MockBase
+
+
+class MockCalledProcessError(Exception):
+    """
+    A mock of subprocess.CalledProcessError
+    """
+
+    def __init__(self):
+        self.returncode = 1
+        self.cmd = ["dummy command"]
+        self.output = "mock output"
+
+
+class MockPerfAnalyzerMethods(MockBase):
+    """
+    Mocks the subprocess module functions used in
+    model_analyzer/perf_analyzer/perf_analyzer.py
+    Provides functions to check operation.
+    """
+
+    def __init__(self):
+        self.mock_popen = MagicMock()
+        self.mock_popen.pid = 10
+        self.mock_popen.returncode = 0
+
+        self.mock_popen_constructor = MagicMock()
+        self.mock_popen_constructor.return_value = self.mock_popen
+
+        self.patcher_popen_stdout_read = patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.Popen",
+            self.mock_popen_constructor,
+        )
+
+        self.mock_file = MagicMock()
+        self.mock_file.read.return_value = b""
+        self.patcher_open = patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.tempfile.NamedTemporaryFile",
+            Mock(return_value=self.mock_file),
+        )
+        super().__init__()
+        self._fill_patchers()
+
+    def start(self):
+        """
+        Start the patchers
+        """
+
+        self.popen_stdout_read = self.patcher_popen_stdout_read.start()
+        self.open_mock = self.patcher_open.start()
+
+    def _fill_patchers(self):
+        """
+        Fills patcher list
+        """
+        self._patchers.append(self.patcher_open)
+        self._patchers.append(self.patcher_popen_stdout_read)
+
+    def set_perf_analyzer_result_string(self, output_string):
+        """
+        Sets the return value of mock_file
+        """
+
+        self.mock_file.read.return_value = output_string.encode("utf-8")
+
+    def get_perf_analyzer_popen_call_count(self):
+        """
+        Get PerfAnalyzer.Popen call count
+        """
+
+        return self.mock_popen_constructor.call_count
+
+    def set_perf_analyzer_return_code(self, returncode):
+        """
+        Sets the returncode of Popen process
+        """
+
+        self.mock_popen.returncode = returncode
+
+    def reset(self):
+        """
+        Resets the side effects
+        and return values of the
+        mocks in this module
+        """
+        self.mock_file.read.return_value = b""
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_psutil.py b/src/c++/perf_analyzer/tests/mocks/mock_psutil.py
new file mode 100755
index 000000000..feb507932
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_psutil.py
@@ -0,0 +1,35 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, patch
+
+from .mock_base import MockBase
+
+
+class MockPSUtil(MockBase):
+    def __init__(self):
+        super().__init__()
+        self._fill_patchers()
+
+    def _fill_patchers(self):
+        mock_process = MagicMock()
+        mock_process().cpu_percent.return_value = 5
+        self._patchers.append(
+            patch(
+                "model_analyzer.perf_analyzer.perf_analyzer.psutil.Process",
+                mock_process,
+            )
+        )
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_requests.py b/src/c++/perf_analyzer/tests/mocks/mock_requests.py
new file mode 100755
index 000000000..6e9902cc1
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_requests.py
@@ -0,0 +1,54 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import Mock, patch
+
+from .mock_base import MockBase
+
+
+class MockRequests(MockBase):
+    def __init__(self, mock_paths):
+        self._mock_paths = mock_paths
+        response_attrs = {"content": Mock()}
+        request_attrs = {"get": Mock(**response_attrs)}
+        self._patchers_requests = {}
+        self._request_mocks = {}
+        for path in mock_paths:
+            self._patchers_requests[path] = patch(
+                f"{path}.requests", Mock(**request_attrs)
+            )
+        super().__init__()
+        self._fill_patchers()
+
+    def _fill_patchers(self):
+        """
+        Fills the patcher list for destruction
+        """
+
+        for patcher in self._patchers_requests.values():
+            self._patchers.append(patcher)
+
+    def start(self):
+        """
+        start the patchers
+        """
+
+        for path in self._mock_paths:
+            self._request_mocks[path] = self._patchers_requests[path].start()
+
+    def set_get_request_response(self, response):
+        for mock in self._request_mocks.values():
+            mock.get.return_value.content = response
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_run_configs.py b/src/c++/perf_analyzer/tests/mocks/mock_run_configs.py
new file mode 100755
index 000000000..dc3b37a00
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_run_configs.py
@@ -0,0 +1,144 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import itertools
+
+
+class MockRunConfig:
+    """
+    Mock class that only contains the important values from the model_config and
+    perf_config from a ModelRunConfig object
+    """
+
+    def load_from_model_run_config(self, model_run_config):
+        """Populate from a ModelRunConfig object"""
+        model_config = model_run_config.model_config().get_config()
+        perf_config = model_run_config.perf_config()
+        self._load_from_model_and_perf_config(model_config, perf_config)
+
+    def load_from_dict(self, dict):
+        """Populate from a dictionary"""
+        self._config = dict
+
+    def get_config_tuple(self):
+        """
+        Convert and return the data from a dict into a tuple
+        of (key,value,key,value,...) where the keys are in a
+        sorted order
+        """
+        list_of_key_value_pairs = [
+            (y, self._config[y])
+            for y in sorted(self._config.keys())
+            if self._config[y] != None
+        ]
+        out_tuple = tuple(item for x in list_of_key_value_pairs for item in x)
+        return out_tuple
+
+    def _load_from_model_and_perf_config(self, model_config, perf_config):
+        """
+        Given a model configuration and perf configuration, extract and store
+        the key bits of information
+        """
+        max_batch_size = None
+        if model_config.get("max_batch_size") is not None:
+            max_batch_size = model_config["max_batch_size"]
+
+        instances = None
+        kind = None
+        if model_config.get("instance_group") is not None:
+            instances = model_config["instance_group"][0]["count"]
+            kind = model_config["instance_group"][0]["kind"]
+
+        dynamic_batching = None
+        max_queue_delay = None
+        if model_config.get("dynamic_batching") is not None:
+            dynamic_batching = 0
+            max_queue_delay = model_config["dynamic_batching"].get(
+                "max_queue_delay_microseconds"
+            )
+
+        batch_size = perf_config.__getitem__("batch-size")
+        concurrency = perf_config.__getitem__("concurrency-range")
+
+        self._config = {
+            "kind": kind,
+            "batch_sizes": batch_size,
+            "batching": dynamic_batching,
+            "concurrency": concurrency,
+            "instances": instances,
+            "max_batch_size": max_batch_size,
+            "max_queue_delay": max_queue_delay,
+        }
+
+
+class MockRunConfigs:
+    """
+    This class holds a list of MockRunConfig
+    """
+
+    def __init__(self):
+        self._configs = []
+
+    def get_num_configs(self) -> int:
+        """Returns the number of configs"""
+        return len(self._configs)
+
+    def get_configs_set(self) -> set:
+        """Returns a set of the configs"""
+        configs_set = {config.get_config_tuple() for config in self._configs}
+        return configs_set
+
+    def add_from_model_run_config(self, config):
+        """Add a single config from a ModelRunConfig"""
+
+        mock_run_config = MockRunConfig()
+        mock_run_config.load_from_model_run_config(config)
+        self._configs.append(mock_run_config)
+
+    def add_from_dict(self, config):
+        """Add a single config from a dict"""
+
+        mock_run_config = MockRunConfig()
+        mock_run_config.load_from_dict(config)
+        self._configs.append(mock_run_config)
+
+    def populate_from_ranges(self, ranges):
+        """
+        Given a dict of key-to-list, create the set of ModelRunConfigs based on
+        the full cartesian product of each dict in the list
+
+        For example, passing in
+        [{ A: [1,2], B: [3,4], C:[0] },  { A: [5], B: [8,9], C: [10] }]
+        would create configs with the following values:
+          - A=1, B=3, C=0
+          - A=1, B=4, C=0
+          - A=2, B=3, C=0
+          - A=2, B=4, C=0
+          - A=5, B=8, C=10
+          - A=5, B=9, C=10
+        """
+        for ranges_dict in ranges:
+            ranges_keys = list(sorted(ranges_dict.keys()))
+            value_lists = []
+            for key in ranges_keys:
+                value_lists.append(ranges_dict[key])
+            configs = list(itertools.product(*value_lists))
+
+            for value in configs:
+                config_dict = {
+                    ranges_keys[i]: value[i] for i in range(len(ranges_keys))
+                }
+                self.add_from_dict(config_dict)
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_server.py b/src/c++/perf_analyzer/tests/mocks/mock_server.py
new file mode 100755
index 000000000..550e6c4ba
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_server.py
@@ -0,0 +1,46 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from abc import abstractmethod
+
+from .mock_base import MockBase
+
+
+class MockServerMethods(MockBase):
+    """
+    Interface for a mock server declaring
+    the methods it must provide.
+    """
+
+    @abstractmethod
+    def assert_server_process_start_called_with(self, *args, **kwargs):
+        """
+        Asserts that the tritonserver process was started with
+        the supplied arguments
+        """
+
+    @abstractmethod
+    def assert_server_process_terminate_called(self):
+        """
+        Assert that the server process was stopped
+        """
+
+    @abstractmethod
+    def assert_cpu_stats_called(self):
+        """
+        Assert that correct cpu memory stats
+        api functions wer called for this implementation
+        """
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_server_docker.py b/src/c++/perf_analyzer/tests/mocks/mock_server_docker.py
new file mode 100755
index 000000000..0c88211f2
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_server_docker.py
@@ -0,0 +1,179 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, Mock, patch
+
+from .mock_api_error import MockAPIError
+from .mock_server import MockServerMethods
+
+
+class MockServerDockerMethods(MockServerMethods):
+    """
+    Mocks the docker module as used in
+    model_analyzer/triton/server/server_docker.py. Provides functions to
+    check operation.
+    """
+
+    TEST_LOG = "Triton Server Test Log"
+    TEST_MEM = "10.0e3"
+
+    def __init__(self):
+        docker_container_attrs = {
+            "exec_run": MagicMock(return_value=(None, bytes(self.TEST_MEM, "utf-8"))),
+            "stats": Mock(
+                return_value={
+                    "memory_stats": {"usage": 0.0, "max_usage": 0.0, "limits": 0.0}
+                }
+            ),
+        }
+        docker_client_attrs = {
+            "containers.run": Mock(return_value=Mock(**docker_container_attrs))
+        }
+        docker_attrs = {
+            "from_env": Mock(return_value=Mock(**docker_client_attrs)),
+            "types.DeviceRequest": Mock(return_value=0),
+        }
+        self.patcher_docker = patch(
+            "model_analyzer.triton.server.server_docker.docker", Mock(**docker_attrs)
+        )
+        super().__init__()
+        self._fill_patchers()
+
+    def start(self):
+        """
+        Start the patchers
+        """
+
+        self.mock = self.patcher_docker.start()
+
+    def _fill_patchers(self):
+        """
+        Fill patcher list
+        """
+
+        self._patchers.append(self.patcher_docker)
+
+    def _assert_docker_initialized(self):
+        """
+        Asserts that docker.from_env
+        was called to initialize
+        docker client
+        """
+
+        self.mock.from_env.assert_called()
+
+    def _assert_docker_exec_run_with_args(self, cmd, stream):
+        """
+        Asserts that a command cmd was run on the docker container
+        with the given stream value
+        """
+
+        self.mock.from_env.return_value.containers.run.return_value.exec_run.assert_any_call(
+            cmd=cmd, stream=stream
+        )
+
+    def assert_server_process_start_called_with(
+        self,
+        cmd,
+        model_repository_path,
+        triton_image,
+        device_requests,
+        gpus,
+        http_port=8000,
+        grpc_port=8001,
+        metrics_port=8002,
+        mounts=None,
+        labels=None,
+        shm_size=None,
+    ):
+        """
+        Asserts that a triton container was created using the
+        supplied arguments
+        """
+
+        if mounts is None:
+            mounts = []
+        if labels is None:
+            labels = {}
+
+        self._assert_docker_initialized()
+
+        env_cmds = [
+            f"CUDA_VISIBLE_DEVICES={','.join([gpu.device_uuid() for gpu in gpus])}"
+        ]
+
+        mock_volumes = {
+            model_repository_path: {"bind": model_repository_path, "mode": "ro"}
+        }
+        for mount_str in mounts:
+            host_path, dest, mode = mount_str.split(":")
+            mock_volumes[host_path] = {"bind": dest, "mode": mode}
+
+        cmd = " ".join(env_cmds + [cmd])
+        mock_ports = {http_port: 8000, grpc_port: 8001, metrics_port: 8002}
+        self.mock.from_env.return_value.containers.run.assert_called_once_with(
+            command=f'bash -c "{cmd}"',
+            init=True,
+            image=triton_image,
+            device_requests=device_requests,
+            volumes=mock_volumes,
+            labels=labels,
+            ports=mock_ports,
+            publish_all_ports=True,
+            tty=False,
+            stdin_open=False,
+            detach=True,
+            shm_size=shm_size,
+        )
+
+    def raise_exception_on_container_run(self):
+        """
+        Raises MockAPIError on container run
+        """
+        self.exception_on_run_patcher = patch(
+            "model_analyzer.triton.server.server_docker.docker.errors.APIError",
+            MockAPIError,
+        )
+        self.api_error_mock = self.exception_on_run_patcher.start()
+        self.mock.from_env.return_value.containers.run.side_effect = self.api_error_mock
+
+    def stop_raise_exception_on_container_run(self):
+        """
+        Stop raising exception on container run
+        """
+        self.mock.from_env.return_value.containers.run.side_effect = None
+        self.exception_on_run_patcher.stop()
+
+    def assert_server_process_terminate_called(self):
+        """
+        Asserts that stop was called on the return value of containers.run
+        """
+
+        self.mock.from_env.return_value.containers.run.return_value.stop.assert_called()
+        self.mock.from_env.return_value.containers.run.return_value.remove.assert_called()
+
+    def assert_cpu_stats_called(self):
+        """
+        Checks the call to docker.Container.stats
+        """
+
+        self._assert_docker_exec_run_with_args(
+            cmd="bash -c \"pmap -x $(pgrep tritonserver) | tail -n1 | awk '{print $4}'\"",
+            stream=False,
+        )
+        self._assert_docker_exec_run_with_args(
+            cmd="bash -c \"free | awk '{if(NR==2)print $7}'\"", stream=False
+        )
diff --git a/src/c++/perf_analyzer/tests/mocks/mock_server_local.py b/src/c++/perf_analyzer/tests/mocks/mock_server_local.py
new file mode 100755
index 000000000..8715fc0af
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/mocks/mock_server_local.py
@@ -0,0 +1,131 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+from unittest.mock import MagicMock, Mock, patch
+
+from .mock_server import MockServerMethods
+
+
+class MockServerLocalMethods(MockServerMethods):
+    """
+    Mocks the subprocess functions used in
+    model_analyzer/triton/server/server_local.py.
+    Provides functions to check operation.
+    """
+
+    def __init__(self):
+        memory_full_attrs = {"uss": 0}
+        virtual_memory_attrs = {"available": 0}
+        process_attrs = {
+            "memory_full_info": Mock(return_value=Mock(**memory_full_attrs))
+        }
+        psutil_attrs = {
+            "Process": Mock(return_value=Mock(**process_attrs)),
+            "virtual_memory": Mock(return_value=Mock(**virtual_memory_attrs)),
+        }
+        Popen_attrs = {
+            "communicate.return_value": ("Triton Server Test Log", "Test Error")
+        }
+        self.patcher_popen = patch(
+            "model_analyzer.triton.server.server_local.Popen",
+            Mock(return_value=Mock(**Popen_attrs)),
+        )
+        self.patcher_stdout = patch(
+            "model_analyzer.triton.server.server_local.STDOUT", MagicMock()
+        )
+        self.patcher_pipe = patch(
+            "model_analyzer.triton.server.server_local.DEVNULL", MagicMock()
+        )
+        self.patcher_psutil = patch(
+            "model_analyzer.triton.server.server_local.psutil", Mock(**psutil_attrs)
+        )
+        super().__init__()
+        self._fill_patchers()
+
+    def start(self):
+        """
+        Start the patchers
+        """
+
+        self.popen_mock = self.patcher_popen.start()
+        self.stdout_mock = self.patcher_stdout.start()
+        self.pipe_mock = self.patcher_pipe.start()
+        self.psutil_mock = self.patcher_psutil.start()
+
+    def _fill_patchers(self):
+        """
+        Fill patcher list
+        """
+
+        self._patchers.append(self.patcher_popen)
+        self._patchers.append(self.patcher_stdout)
+        self._patchers.append(self.patcher_pipe)
+        self._patchers.append(self.patcher_psutil)
+
+    def assert_server_process_start_called_with(self, cmd, gpus, stdout=MagicMock()):
+        """
+        Asserts that Popen was called
+        with the cmd provided.
+        """
+
+        env = os.environ.copy()
+        env["CUDA_VISIBLE_DEVICES"] = ",".join([gpu.device_uuid() for gpu in gpus])
+
+        self.popen_mock.assert_called_once_with(
+            cmd,
+            stdout=stdout,
+            stderr=self.stdout_mock,
+            start_new_session=True,
+            universal_newlines=True,
+            env=env,
+        )
+
+    def assert_server_process_terminate_called(self):
+        """
+        Asserts that terminate was called on
+        the pipe (Popen object).
+        """
+
+        self.popen_mock.return_value.terminate.assert_called()
+        self.popen_mock.return_value.communicate.assert_called()
+
+    def set_cpu_memory_values(self, used_mem, free_mem):
+        """
+        Sets the value of psutil.Process.memory_full_info.uss
+        and psutil.virtual_memory.available
+        """
+
+        self.psutil_mock.Process.return_value.memory_full_info.return_value.uss = (
+            used_mem
+        )
+        self.psutil_mock.virtual_memory.return_value.available = free_mem
+
+    def assert_cpu_stats_called(self):
+        """
+        Checks the call to psutil.Process.memory_full_info and psutil.virtual_memory
+        """
+
+        self.psutil_mock.Process.return_value.memory_full_info.assert_called()
+        self.psutil_mock.virtual_memory.assert_called()
+
+    def assert_cpu_stats_not_called(self):
+        """
+        Checks no calls to psutil.Process.memory_full_info and psutil.virtual_memory
+        """
+
+        self.psutil_mock.Process.return_value.memory_full_info.assert_not_called()
+        self.psutil_mock.virtual_memory.assert_not_called()
diff --git a/src/c++/perf_analyzer/tests/test_analyzer.py b/src/c++/perf_analyzer/tests/test_analyzer.py
new file mode 100755
index 000000000..68680ff57
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_analyzer.py
@@ -0,0 +1,241 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.analyzer import Analyzer
+from model_analyzer.config.input.config_status import ConfigStatus
+from model_analyzer.config.run.model_run_config import ModelRunConfig
+from model_analyzer.config.run.run_config import RunConfig
+from model_analyzer.constants import CONFIG_PARSER_SUCCESS
+from model_analyzer.result.results import Results
+from model_analyzer.result.run_config_result import RunConfigResult
+from model_analyzer.result.sorted_results import SortedResults
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+from model_analyzer.triton.model.model_config import ModelConfig
+from model_analyzer.triton.model.model_config_variant import ModelConfigVariant
+from tests.common.test_utils import evaluate_mock_config
+
+from .common import test_result_collector as trc
+
+
+class TestAnalyzer(trc.TestResultCollector):
+    """
+    Tests the methods of the Analyzer class
+    """
+
+    def tearDown(self):
+        patch.stopall()
+
+    def mock_get_state_variable(self, name):
+        if name == "ResultManager.results":
+            return Results()
+        else:
+            return {
+                "model1": {
+                    "config1": None,
+                    "config2": None,
+                    "config3": None,
+                    "config4": None,
+                }
+            }
+
+    def mock_get_list_of_models(self):
+        return ["model1"]
+
+    @patch.multiple(
+        f"{AnalyzerStateManager.__module__}.AnalyzerStateManager",
+        get_state_variable=mock_get_state_variable,
+        exiting=lambda _: False,
+    )
+    @patch.multiple(
+        f"{Analyzer.__module__}.Analyzer",
+        _create_metrics_manager=MagicMock(),
+        _create_model_manager=MagicMock(),
+        _get_server_only_metrics=MagicMock(),
+        _profile_models=MagicMock(),
+        _check_for_perf_analyzer_errors=MagicMock(),
+    )
+    def test_profile_skip_summary_reports(self, **mocks):
+        """
+        Tests when the skip_summary_reports config option is turned on,
+        the profile stage does not create any summary reports.
+
+        NOTE: this test only ensures that the reports are not created with
+        the default export-path.
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "/tmp",
+            "--profile-models",
+            "model1",
+            "--config-file",
+            "/tmp/my_config.yml",
+            "--checkpoint-directory",
+            "/tmp/my_checkpoints",
+            "--skip-summary-reports",
+        ]
+        config = evaluate_mock_config(args, "", subcommand="profile")
+        state_manager = AnalyzerStateManager(config, None)
+        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
+        analyzer.profile(client=None, gpus=None, mode=None, verbose=False)
+
+        path = os.getcwd()
+        self.assertFalse(os.path.exists(os.path.join(path, "plots")))
+        self.assertFalse(os.path.exists(os.path.join(path, "results")))
+        self.assertFalse(os.path.exists(os.path.join(path, "reports")))
+
+    def mock_top_n_results(self, model_name=None, n=SortedResults.GET_ALL_RESULTS):
+        rc1 = RunConfig({})
+        rc1.add_model_run_config(
+            ModelRunConfig(
+                "fake_model_name",
+                ModelConfigVariant(
+                    ModelConfig.create_from_dictionary({"name": "config1"}),
+                    "config1",
+                ),
+                MagicMock(),
+            )
+        )
+        rc2 = RunConfig({})
+        rc2.add_model_run_config(
+            ModelRunConfig(
+                "fake_model_name",
+                ModelConfigVariant(
+                    ModelConfig.create_from_dictionary({"name": "config3"}),
+                    "config3",
+                ),
+                MagicMock(),
+            )
+        )
+        rc3 = RunConfig({})
+        rc3.add_model_run_config(
+            ModelRunConfig(
+                "fake_model_name",
+                ModelConfigVariant(
+                    ModelConfig.create_from_dictionary({"name": "config4"}),
+                    "config4",
+                ),
+                MagicMock(),
+            )
+        )
+
+        return [
+            RunConfigResult("fake_model_name", rc1, MagicMock(), MagicMock()),
+            RunConfigResult("fake_model_name", rc2, MagicMock(), MagicMock()),
+            RunConfigResult("fake_model_name", rc3, MagicMock(), MagicMock()),
+        ]
+
+    @patch(
+        "model_analyzer.config.input.config_command_profile.file_path_validator",
+        lambda _: ConfigStatus(status=CONFIG_PARSER_SUCCESS),
+    )
+    @patch(
+        "model_analyzer.config.input.config_command_profile.ConfigCommandProfile._preprocess_and_verify_arguments",
+        lambda _: None,
+    )
+    @patch(
+        "model_analyzer.result.result_manager.ResultManager.top_n_results",
+        mock_top_n_results,
+    )
+    def test_get_report_command_help_string(self):
+        """
+        Tests that the member function returning the report command help string
+        works correctly.
+        """
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "--profile-models",
+            "model1",
+            "--config-file",
+            "/tmp/my_config.yml",
+            "--checkpoint-directory",
+            "/tmp/my_checkpoints",
+            "--export-path",
+            "/tmp/my_export_path",
+            "--model-repository",
+            "cli_repository",
+        ]
+        config = evaluate_mock_config(args, "", subcommand="profile")
+        state_manager = AnalyzerStateManager(config, None)
+        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
+        self.assertEqual(
+            analyzer._get_report_command_help_string(model_name="model1"),
+            "To generate detailed reports for the 3 best model1 configurations, run "
+            "`model-analyzer report --report-model-configs "
+            "config1,config3,config4 --export-path /tmp/my_export_path "
+            "--config-file /tmp/my_config.yml --checkpoint-directory "
+            "/tmp/my_checkpoints`",
+        )
+
+    def test_multiple_models_in_report_model_config(self):
+        """
+        Test that multiple different models in the report are detected
+        """
+        # Single model config
+        args = [
+            "model-analyzer",
+            "report",
+            "--report-model-configs",
+            "modelA_config_0",
+            "--checkpoint-directory",
+            "/tmp/my_checkpoints",
+        ]
+        config = evaluate_mock_config(args, "", subcommand="report")
+        state_manager = AnalyzerStateManager(config, None)
+        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
+
+        self.assertFalse(analyzer._multiple_models_in_report_model_config())
+
+        # Multiple different models
+        args = [
+            "model-analyzer",
+            "report",
+            "--report-model-configs",
+            "modelA_config_0,modelB_config_1",
+            "--checkpoint-directory",
+            "/tmp/my_checkpoints",
+        ]
+        config = evaluate_mock_config(args, "", subcommand="report")
+        state_manager = AnalyzerStateManager(config, None)
+        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
+
+        self.assertTrue(analyzer._multiple_models_in_report_model_config())
+
+        # Single model - multiple configs
+        args = [
+            "model-analyzer",
+            "report",
+            "--report-model-configs",
+            "modelA_config_0,modelA_config_1",
+            "--checkpoint-directory",
+            "/tmp/my_checkpoints",
+        ]
+        config = evaluate_mock_config(args, "", subcommand="report")
+        state_manager = AnalyzerStateManager(config, None)
+        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
+
+        self.assertFalse(analyzer._multiple_models_in_report_model_config())
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_bls_report_manager.py b/src/c++/perf_analyzer/tests/test_bls_report_manager.py
new file mode 100755
index 000000000..911369081
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_bls_report_manager.py
@@ -0,0 +1,172 @@
+#!/usr/bin/env python3
+
+# Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import shutil
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.reports.report_manager import ReportManager
+from model_analyzer.result.result_manager import ResultManager
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+
+from .common import test_result_collector as trc
+from .common.test_utils import ROOT_DIR, evaluate_mock_config
+
+
+class TestBLSReportManager(trc.TestResultCollector):
+    def _init_managers(
+        self,
+        models="test_model",
+        num_configs_per_model=10,
+        mode="online",
+        subcommand="profile",
+    ):
+        args = ["model-analyzer", subcommand, "-f", "path-to-config-file"]
+        args.extend(["--checkpoint-directory", f"{ROOT_DIR}/bls-ckpt"])
+
+        if subcommand == "profile":
+            args.extend(["--profile-models", models])
+            args.extend(["--model-repository", "/tmp"])
+        else:
+            args.extend(["--report-model-configs", models])
+
+        yaml_str = f"""
+            client_protocol: grpc
+            export_path: {ROOT_DIR}
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand=subcommand)
+        state_manager = AnalyzerStateManager(config=config, server=None)
+
+        state_manager.load_checkpoint(checkpoint_required=True)
+        gpu_info = {"gpu_uuid": {"name": "fake_gpu_name", "total_memory": 1024000000}}
+        self.result_manager = ResultManager(
+            config=config, state_manager=state_manager, constraint_manager=MagicMock()
+        )
+        self.report_manager = ReportManager(
+            mode=mode,
+            config=config,
+            gpu_info=gpu_info,
+            result_manager=self.result_manager,
+            constraint_manager=MagicMock(),
+        )
+
+    def test_bls_summary(self):
+        """
+        Ensures the summary report sentence and table are accurate for a basic bls model (loaded from a checkpoint)
+        """
+        self._init_managers(models="bls", subcommand="profile")
+
+        self.report_manager.create_summaries()
+
+        expected_summary_sentence = (
+            "In 54 measurements across 9 configurations, <strong>bls_config_6</strong> "
+            "is <strong>1111%</strong> better than the default configuration at maximizing throughput, "
+            "under the given constraints, on GPU(s) TITAN RTX.<UL><LI> <strong>bls_config_6</strong>: 8 GPU instances with a max batch size of 0 on platform python "
+            "</LI><BR>Which is comprised of the following composing models: "
+            "<UL><LI> <strong>add_config_4</strong>: 6 GPU instances with a max batch size of 0 on platform python "
+            "</LI><LI> <strong>sub_config_0</strong>: 1 GPU instance with a max batch size of 0 on platform python </LI> </UL>"
+        )
+
+        summary_table, summary_sentence = self.report_manager._build_summary_table(
+            report_key="bls",
+            num_measurements=54,
+            num_configurations=9,
+            gpu_name="TITAN RTX",
+            report_gpu_metrics=True,
+        )
+
+        self.assertEqual(summary_sentence, expected_summary_sentence)
+
+        # Check the first row of the table
+        self.assertEqual(summary_table._rows[0][0], "bls_config_6")  # model config name
+        self.assertEqual(summary_table._rows[0][1], "(0, 0)")  # max batch size
+        self.assertEqual(
+            summary_table._rows[0][2], "(Disabled, Disabled)"
+        )  # dynamic batching
+        self.assertEqual(summary_table._rows[0][3], "(6:GPU, 1:GPU)")  # instance count
+        self.assertEqual(summary_table._rows[0][4], "2.215")  # p99 latency
+        self.assertEqual(summary_table._rows[0][5], "10077.3")  # throughput
+        self.assertEqual(summary_table._rows[0][6], 874)  # max gpu memory
+        self.assertEqual(summary_table._rows[0][7], 0.0)  # GPU utilization
+
+    def test_bls_summary_cpu_only(self):
+        """
+        Ensures the summary report sentence and table are accurate for a basic BLS model (loaded from a checkpoint)
+        when the cpu only flag is set
+        """
+
+        self._init_managers(models="bls", subcommand="profile")
+
+        self.report_manager.create_summaries()
+
+        expected_summary_sentence = (
+            "In 54 measurements across 9 configurations, <strong>bls_config_6</strong> "
+            "is <strong>1111%</strong> better than the default configuration at maximizing throughput, "
+            "under the given constraints.<UL><LI> <strong>bls_config_6</strong>: 8 GPU instances with a max batch size of 0 on platform python "
+            "</LI><BR>Which is comprised of the following composing models: "
+            "<UL><LI> <strong>add_config_4</strong>: 6 GPU instances with a max batch size of 0 on platform python "
+            "</LI><LI> <strong>sub_config_0</strong>: 1 GPU instance with a max batch size of 0 on platform python </LI> </UL>"
+        )
+
+        summary_table, summary_sentence = self.report_manager._build_summary_table(
+            report_key="bls",
+            num_measurements=54,
+            num_configurations=9,
+            gpu_name="TITAN RTX",
+            report_gpu_metrics=False,
+        )
+
+        self.assertEqual(summary_sentence, expected_summary_sentence)
+
+        # Check the first row of the table
+        self.assertEqual(summary_table._rows[0][0], "bls_config_6")  # model config name
+        self.assertEqual(summary_table._rows[0][1], "0, 0")  # max batch size
+        self.assertEqual(
+            summary_table._rows[0][2], "(Disabled, Disabled)"
+        )  # dynamic batching
+        self.assertEqual(summary_table._rows[0][3], "6:GPU, 1:GPU")  # instance count
+        self.assertEqual(summary_table._rows[0][4], "2.215")  # p99 latency
+        self.assertEqual(summary_table._rows[0][5], "10077.3")  # throughput
+
+    def test_bls_detailed(self):
+        """
+        Ensures the detailed report sentence is accurate for a BLS model (loaded from a checkpoint)
+        """
+        self._init_managers(models="bls_config_6", subcommand="report")
+
+        self.report_manager._add_detailed_report_data()
+        self.report_manager._build_detailed_table("bls_config_6")
+        detailed_sentence = self.report_manager._build_detailed_info("bls_config_6")
+
+        expected_detailed_sentence = (
+            "<strong>bls_config_6</strong> is comprised of the "
+            "following composing models:<LI> <strong>add_config_4</strong>: "
+            "6 GPU instances with a max batch size of 0 on platform python </LI><LI> "
+            "<strong>sub_config_0</strong>: 1 GPU instance with a max batch size of 0 "
+            "on platform python </LI><br>12 measurement(s) "
+            "were obtained for the model config on GPU(s)  with total memory 0 GB."
+        )
+
+        self.assertEqual(detailed_sentence, expected_detailed_sentence)
+
+    def tearDown(self):
+        patch.stopall()
+        shutil.rmtree(f"{ROOT_DIR}/reports")
+        shutil.rmtree(f"{ROOT_DIR}/plots")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_cli.py b/src/c++/perf_analyzer/tests/test_cli.py
new file mode 100755
index 000000000..98ec60237
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_cli.py
@@ -0,0 +1,604 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# This test exercises all of the cli options.
+# It does a basic test of the cli and help message.
+# Then the rest of the testing uses an OptionStruct, which holds all
+# of the data necessary to test a command, and feeds that to the
+# CLI parser. The result of the CLI parsing is compared against the
+# expected value for the CLI. Default values are also verified as well as
+# values that are expected to cause failures.
+
+import copy
+import os
+import sys
+import unittest
+from unittest.mock import MagicMock, patch
+
+import psutil
+
+from model_analyzer.cli.cli import CLI
+from model_analyzer.config.input.config_command_profile import ConfigCommandProfile
+from model_analyzer.config.input.config_command_report import ConfigCommandReport
+from model_analyzer.config.input.config_defaults import DEFAULT_TRITON_DOCKER_IMAGE
+from model_analyzer.config.input.config_status import ConfigStatus
+from model_analyzer.constants import CONFIG_PARSER_SUCCESS
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+
+from .common import test_result_collector as trc
+
+
+def tearDown(self):
+    patch.stopall()
+
+
+def get_test_options():
+    """
+    Returns the list of OptionStructs that are used for testing.
+    """
+    # yapf: disable
+    options = [
+        #Boolean options
+        # Options format:
+        #   (bool, MA step, long_option)
+        OptionStruct("bool", "profile","--override-output-model-repository"),
+        OptionStruct("bool", "profile","--collect-cpu-metrics"),
+        OptionStruct("bool", "profile","--perf-output"),
+        OptionStruct("bool", "profile","--run-config-search-disable"),
+        OptionStruct("bool", "profile","--run-config-profile-models-concurrently-enable"),
+        OptionStruct("bool", "profile","--request-rate-search-enable"),
+        OptionStruct("bool", "profile","--reload-model-disable"),
+        OptionStruct("bool", "profile","--early-exit-enable"),
+        OptionStruct("bool", "profile","--skip-summary-reports"),
+        OptionStruct("bool", "profile","--skip-detailed-reports"),
+        OptionStruct("bool", "profile","--always-report-gpu-metrics"),
+        #Int/Float options
+        # Options format:
+        #   (int/float, MA step, long_option, short_option, test_value, expected_default_value)
+        # The following options can be None:
+        #   short_option
+        #   expected_default_value
+        OptionStruct("int", "profile", "--client-max-retries", "-r", "125", "50"),
+        OptionStruct("int", "profile", "--duration-seconds", "-d", "10", "3"),
+        OptionStruct("int", "profile", "--perf-analyzer-timeout", None, "100", "600"),
+        OptionStruct("int", "profile", "--perf-analyzer-max-auto-adjusts", None, "100", "10"),
+        OptionStruct("int", "profile", "--run-config-search-min-concurrency", None, "2", "1"),
+        OptionStruct("int", "profile", "--run-config-search-max-concurrency", None, "100", "1024"),
+        OptionStruct("int", "profile", "--run-config-search-min-request-rate", None, "2", "16"),
+        OptionStruct("int", "profile", "--run-config-search-max-request-rate", None, "100", "8192"),
+        OptionStruct("int", "profile", "--run-config-search-min-model-batch-size", None, "100", "1"),
+        OptionStruct("int", "profile", "--run-config-search-max-model-batch-size", None, "100", "128"),
+        OptionStruct("int", "profile", "--run-config-search-min-instance-count", None, "2", "1"),
+        OptionStruct("int", "profile", "--run-config-search-max-instance-count", None, "10", "5"),
+        OptionStruct("int", "profile", "--run-config-search-max-binary-search-steps", None, "10", "5"),
+        OptionStruct("float", "profile", "--monitoring-interval", "-i", "10.0", "1.0"),
+        OptionStruct("float", "profile", "--perf-analyzer-cpu-util", None, "10.0", str(psutil.cpu_count() * 80.0)),
+        OptionStruct("int", "profile", "--num-configs-per-model", None, "10", "3"),
+        OptionStruct("int", "profile", "--num-top-model-configs", None, "10", "0"),
+        OptionStruct("int", "profile", "--latency-budget", None, "200", None),
+        OptionStruct("int", "profile", "--min-throughput", None, "300", None),
+
+        #String options
+        # Options format:
+        #   (string, MA step, long_flag, short_flag, test_value, expected_default_value, expected_failing_value, extra_commands)
+        # The following options can be None:
+        #   short_flag
+        #   expected_default_value
+        #   expected_failing_value
+        # For options with choices, list the test_values in a list of strings
+        OptionStruct("string", "profile", "--config-file", "-f", "baz", None, None),
+        OptionStruct("string", "profile", "--checkpoint-directory", "-s", "./test_dir", os.path.join(os.getcwd(), "checkpoints"), None),
+        OptionStruct("string", "profile", "--output-model-repository-path", None, "./test_dir", os.path.join(os.getcwd(), "output_model_repository"), None),
+        OptionStruct("string", "profile", "--client-protocol", None, ["http", "grpc"], "grpc", "SHOULD_FAIL"),
+        OptionStruct("string", "profile", "--perf-analyzer-path", None, ".", "perf_analyzer", None),
+        OptionStruct("string", "profile", "--perf-output-path", None, ".", None, None),
+        OptionStruct("string", "profile", "--triton-docker-image", None, "test_image", DEFAULT_TRITON_DOCKER_IMAGE, None),
+        OptionStruct("string", "profile", "--triton-http-endpoint", None, "localhost:4000", "localhost:8000", None),
+        OptionStruct("string", "profile", "--triton-grpc-endpoint", None, "localhost:4001", "localhost:8001", None),
+        OptionStruct("string", "profile", "--triton-metrics-url", None, "localhost:4002", "http://localhost:8002/metrics", None),
+        OptionStruct("string", "profile", "--triton-server-path", None, "test_path", "tritonserver", None),
+        OptionStruct("string", "profile", "--triton-output-path", None, "test_path", None, None),
+        OptionStruct("string", "profile", "--triton-launch-mode", None, ["local", "docker", "remote","c_api"], "local", "SHOULD_FAIL"),
+        OptionStruct("string", "profile", "--triton-install-path", None, "test_path", "/opt/tritonserver", None),
+        OptionStruct("string", "profile", "--checkpoint-directory", "-s", "./test_dir", os.path.join(os.getcwd(), "checkpoints"), None),
+        OptionStruct("string", "profile", "--export-path", "-e", "./test_dir", os.getcwd(), None),
+        OptionStruct("string", "profile", "--filename-model-inference", None, "foo", "metrics-model-inference.csv", None),
+        OptionStruct("string", "profile", "--filename-model-gpu", None, "foo", "metrics-model-gpu.csv", None),
+        OptionStruct("string", "profile", "--filename-server-only", None, "foo", "metrics-server-only.csv", None),
+        OptionStruct("string", "profile", "--config-file", "-f", "baz", None, None),
+
+        OptionStruct("string", "report", "--checkpoint-directory", "-s", "./test_dir", os.path.join(os.getcwd(), "checkpoints"), None),
+        OptionStruct("string", "report", "--export-path", "-e", "./test_dir", os.getcwd(), None),
+        OptionStruct("string", "report", "--config-file", "-f", "baz", None, None),
+        OptionStruct("string", "profile", "--triton-docker-shm-size", None, "1G", None, extra_commands=["--triton-launch-mode", "docker"]),
+        OptionStruct("string", "profile","--run-config-search-mode", None, ["quick", "brute"], "brute", "SHOULD_FAIL"),
+
+        #List Options:
+        # Options format:
+        #   (intlist/stringlist, MA step, long_flag, short_flag, test_value, expected_default_value, extra_commands)
+        # The following options can be None:
+        #   short_flag
+        #   expected_default_value
+        OptionStruct("intlist", "profile", "--batch-sizes", "-b", "2, 4, 6", "1"),
+        OptionStruct("intlist", "profile", "--concurrency", "-c", "1, 2, 3", None),
+        OptionStruct("intlist", "profile", "--request-rate", None, "1, 2, 3", None),
+        OptionStruct("stringlist", "profile", "--triton-docker-mounts", None, "a:b:c, d:e:f", None, extra_commands=["--triton-launch-mode", "docker"]),
+        OptionStruct("stringlist", "profile", "--gpus", None, "a, b, c", "all"),
+        OptionStruct("stringlist", "profile", "--inference-output-fields", None, "a, b, c",
+            "model_name,batch_size,concurrency,model_config_path,instance_group,max_batch_size,satisfies_constraints,perf_throughput,perf_latency_p99"),
+        OptionStruct("stringlist", "profile", "--gpu-output-fields", None, "a, b, c",
+            "model_name,gpu_uuid,batch_size,concurrency,model_config_path,instance_group,satisfies_constraints,gpu_used_memory,gpu_utilization,gpu_power_usage"),
+        OptionStruct("stringlist", "profile", "--server-output-fields", None, "a, b, c",
+            "model_name,gpu_uuid,gpu_used_memory,gpu_utilization,gpu_power_usage"),
+
+        # No OP Options:
+        # Option format:
+        # (noop, any MA step, long_flag)
+        # These commands aren't tested directly but are here to ensure that
+        # the count is correct for all options in the config.
+        # Some of these are required to run the subcommand
+        # Others are yaml only options
+        OptionStruct("noop", "profile", "--model-repository"),
+        OptionStruct("noop", "profile", "--profile-models"),
+        OptionStruct("noop", "profile", "--bls-composing-models"),
+        OptionStruct("noop", "profile", "--cpu-only-composing-models"),
+
+        OptionStruct("noop", "report", "--report-model-configs"),
+        OptionStruct("noop", "report", "--output-formats", "-o", ["pdf", "csv", "png"], "pdf", "SHOULD_FAIL"),
+        OptionStruct("noop", "yaml_profile", "constraints"),
+        OptionStruct("noop", "yaml_profile", "objectives"),
+        OptionStruct("noop", "yaml_profile", "weighting"),
+        OptionStruct("noop", "yaml_profile", "triton_server_flags"),
+        OptionStruct("noop", "yaml_profile", "perf_analyzer_flags"),
+        OptionStruct("noop", "yaml_profile", "triton_docker_labels"),
+        OptionStruct("noop", "yaml_profile", "triton_server_environment"),
+        OptionStruct("noop", "yaml_profile", "triton_docker_args"),
+        OptionStruct("noop", "yaml_profile", "plots")
+    ]
+    # yapf: enable
+    return options
+
+
+class CLISubclass(CLI):
+    """
+    Subclass of CLI to overwrite the parse method.
+    Parse takes a list of arguments instead of getting the args
+    from sys.argv
+    """
+
+    def __init__(self):
+        super().__init__()
+
+    def parse(self, parsed_commands=None):
+        args = self._parser.parse_args(parsed_commands[1:])
+        if args.subcommand is None:
+            self._parser.print_help()
+            self._parser.exit()
+        config = self._subcommand_configs[args.subcommand]
+        config.set_config_values(args)
+        return args, config
+
+
+class CLIConfigProfileStruct:
+    """
+    Struct class to hold the common variables shared between profile tests
+    """
+
+    def __init__(self):
+        # yapf: disable
+        self.args = [
+            '/usr/local/bin/model-analyzer',
+            'profile',
+            '--model-repository',
+            'foo',
+            '--profile-models',
+            'bar'
+        ]
+        # yapf: enable
+        config_profile = ConfigCommandProfile()
+        self.cli = CLISubclass()
+        self.cli.add_subcommand(cmd="profile", help="", config=config_profile)
+
+    def parse(self):
+        return self.cli.parse(self.args)
+
+
+class CLIConfigReportStruct:
+    """
+    Struct class to hold the common variables shared between report tests
+    """
+
+    def __init__(self):
+        # yapf: disable
+        self.args = [
+            '/usr/local/bin/model-analyzer',
+            'report',
+            '--report-model-configs',
+            'a, b, c'
+        ]
+        # yapf: enable
+        config_report = ConfigCommandReport()
+        self.cli = CLISubclass()
+        self.cli.add_subcommand(cmd="report", help="", config=config_report)
+
+    def parse(self):
+        return self.cli.parse(self.args)
+
+
+class OptionStruct:
+    """
+    Struct that holds all of the data necessary to test a single command
+    """
+
+    def __init__(
+        self,
+        type,
+        stage,
+        long_flag,
+        short_flag=None,
+        expected_value=None,
+        expected_default_value=None,
+        expected_failing_value=None,
+        extra_commands=None,
+    ):
+        self.long_flag = long_flag
+        self.short_flag = short_flag
+        self.expected_value = expected_value
+        self.expected_default_value = expected_default_value
+        self.expected_failing_value = expected_failing_value
+        self.type = type
+        self.extra_commands = extra_commands
+
+        if stage == "profile":
+            self.cli_subcommand = CLIConfigProfileStruct
+        elif stage == "report":
+            self.cli_subcommand = CLIConfigReportStruct
+
+
+@patch(
+    "model_analyzer.config.input.config_command_profile.file_path_validator",
+    lambda _: ConfigStatus(status=CONFIG_PARSER_SUCCESS),
+)
+@patch(
+    "model_analyzer.config.input.config_command_profile.binary_path_validator",
+    lambda _: ConfigStatus(status=CONFIG_PARSER_SUCCESS),
+)
+@patch(
+    "model_analyzer.config.input.config_command_report.file_path_validator",
+    lambda _: ConfigStatus(status=CONFIG_PARSER_SUCCESS),
+)
+class TestCLI(trc.TestResultCollector):
+    """
+    Tests the methods of the CLI class
+    """
+
+    @patch("model_analyzer.cli.cli.ArgumentParser.print_help")
+    def test_help_message_no_args(self, mock_print_help):
+        """
+        Tests that model-analyzer prints the help message when no arguments are
+        given
+        """
+
+        sys.argv = ["/usr/local/bin/model-analyzer"]
+
+        cli = CLI()
+
+        self.assertRaises(SystemExit, cli.parse)
+        mock_print_help.assert_called()
+
+    def test_basic_cli_config_profile_options(self):
+        """
+        Test the minimal set of cli commands necessary to run Model Analyzer profile
+        """
+        cli = CLIConfigProfileStruct()
+        _, config = cli.parse()
+        model_repo = config.model_repository
+        profile_model = config.profile_models[0].model_name()
+        self.assertEqual("foo", model_repo)
+        self.assertEqual("bar", profile_model)
+
+    @patch(
+        "model_analyzer.config.input.config_command_report.ConfigCommandReport._load_config_file",
+        MagicMock(),
+    )
+    @patch(
+        "model_analyzer.config.input.config_command_report.ConfigCommandReport._preprocess_and_verify_arguments",
+        MagicMock(),
+    )
+    @patch(
+        "model_analyzer.config.input.config_command_profile.ConfigCommandProfile._preprocess_and_verify_arguments",
+        MagicMock(),
+    )
+    @patch(
+        "model_analyzer.config.input.config_command_profile.ConfigCommandProfile._load_config_file",
+        MagicMock(),
+    )
+    @patch(
+        "model_analyzer.config.input.config_command.ConfigCommand._check_for_illegal_config_settings",
+        MagicMock(),
+    )
+    def test_all_options(self):
+        options = get_test_options()
+        all_tested_options_set = set()
+
+        for option in options:
+            all_tested_options_set.add(
+                self._convert_flag_to_use_underscores(option.long_flag)
+            )
+
+            if option.type in ["bool"]:
+                self._test_boolean_option(option)
+            elif option.type in ["int", "float"]:
+                self._test_numeric_option(option)
+            elif option.type in ["string"]:
+                self._test_string_option(option)
+            elif option.type in ["intlist", "stringlist"]:
+                self._test_list_option(option)
+            elif option.type in ["noop"]:
+                pass
+            else:
+                raise (TritonModelAnalyzerException("Invalid option type"))
+
+        self._verify_all_options_tested(all_tested_options_set)
+
+    def _verify_all_options_tested(self, all_tested_options_set):
+        cli_option_set = set()
+
+        # Get all of the options in the CLI Configs
+        structs = [CLIConfigProfileStruct, CLIConfigReportStruct]
+        for struct in structs:
+            cli = struct()
+            _, config = cli.parse()
+            for key in config.get_config().keys():
+                cli_option_set.add(key)
+
+        self.assertEqual(
+            cli_option_set,
+            all_tested_options_set,
+            "The available options for the CLI does not match the available options tested. "
+            "If you recently added or removed a CLI option, please update the OptionStruct list above.",
+        )
+
+    def _test_boolean_option(self, option_struct):
+        option = option_struct.long_flag
+        option_with_underscores = self._convert_flag_to_use_underscores(option)
+        cli = option_struct.cli_subcommand()
+        _, config = cli.parse()
+        option_value = config.get_config().get(option_with_underscores).value()
+        self.assertEqual(option_value, False)
+
+        # Test boolean option
+        cli = option_struct.cli_subcommand()
+        cli.args.extend([option])
+        _, config = cli.parse()
+        option_value = config.get_config().get(option_with_underscores).value()
+        self.assertEqual(option_value, True)
+
+        # Test boolean option followed by value fails
+        cli = option_struct.cli_subcommand()
+        cli.args.extend([option, "SHOULD_FAIL"])
+        with self.assertRaises(SystemExit):
+            _, config = cli.parse()
+
+    def _test_numeric_option(self, option_struct):
+        long_option = option_struct.long_flag
+        short_option = option_struct.short_flag
+        expected_value_string = option_struct.expected_value
+        expected_value = self._convert_string_to_numeric(option_struct.expected_value)
+        expected_default_value = (
+            None
+            if option_struct.expected_default_value is None
+            else self._convert_string_to_numeric(option_struct.expected_default_value)
+        )
+
+        long_option_with_underscores = self._convert_flag_to_use_underscores(
+            long_option
+        )
+
+        self._test_long_flag(
+            long_option,
+            option_struct.cli_subcommand,
+            expected_value_string,
+            long_option_with_underscores,
+            expected_value,
+        )
+
+        self._test_short_flag(
+            short_option,
+            option_struct.cli_subcommand,
+            expected_value_string,
+            long_option_with_underscores,
+            expected_value,
+        )
+
+        if expected_default_value is not None:
+            self._test_expected_default_value(
+                option_struct.cli_subcommand,
+                long_option_with_underscores,
+                expected_default_value,
+            )
+
+    def _test_string_option(self, option_struct):
+        long_option = option_struct.long_flag
+        short_option = option_struct.short_flag
+        expected_value = option_struct.expected_value
+        expected_default_value = option_struct.expected_default_value
+        expected_failing_value = option_struct.expected_failing_value
+        extra_commands = option_struct.extra_commands
+
+        # This covers strings that have choices
+        # Recursively call this method with choices
+        if type(expected_value) is list:
+            for value in expected_value:
+                new_struct = copy.deepcopy(option_struct)
+                new_struct.expected_value = value
+                self._test_string_option(new_struct)
+        else:
+            long_option_with_underscores = self._convert_flag_to_use_underscores(
+                long_option
+            )
+
+            self._test_long_flag(
+                long_option,
+                option_struct.cli_subcommand,
+                expected_value,
+                long_option_with_underscores,
+                expected_value,
+                extra_commands,
+            )
+
+            self._test_short_flag(
+                short_option,
+                option_struct.cli_subcommand,
+                expected_value,
+                long_option_with_underscores,
+                expected_value,
+            )
+
+            if expected_default_value is not None:
+                self._test_expected_default_value(
+                    option_struct.cli_subcommand,
+                    long_option_with_underscores,
+                    expected_default_value,
+                )
+
+            # Verify that a incorrect value causes a failure
+            if expected_failing_value is not None:
+                cli = option_struct.cli_subcommand()
+                cli.args.extend([long_option, expected_failing_value])
+                with self.assertRaises(SystemExit):
+                    _, config = cli.parse()
+
+    def _test_list_option(self, option_struct):
+        long_option = option_struct.long_flag
+        short_option = option_struct.short_flag
+        expected_value = option_struct.expected_value
+        expected_default_value = option_struct.expected_default_value
+        extra_commands = option_struct.extra_commands
+
+        # Convert expected and default values to proper types for comparison
+        if option_struct.type == "intlist":
+            expected_value_converted = self._convert_string_to_int_list(expected_value)
+            if expected_default_value is not None:
+                expected_default_value_converted = self._convert_string_to_int_list(
+                    expected_default_value
+                )
+            else:
+                expected_default_value_converted = None
+        else:
+            expected_value_converted = self._convert_string_to_string_list(
+                expected_value
+            )
+            if expected_default_value is not None:
+                expected_default_value_converted = self._convert_string_to_string_list(
+                    expected_default_value
+                )
+            else:
+                expected_default_value_converted = None
+
+        long_option_with_underscores = self._convert_flag_to_use_underscores(
+            long_option
+        )
+
+        self._test_long_flag(
+            long_option,
+            option_struct.cli_subcommand,
+            expected_value,
+            long_option_with_underscores,
+            expected_value_converted,
+            extra_commands,
+        )
+
+        self._test_short_flag(
+            short_option,
+            option_struct.cli_subcommand,
+            expected_value,
+            long_option_with_underscores,
+            expected_value_converted,
+        )
+
+        if expected_default_value is not None:
+            self._test_expected_default_value(
+                option_struct.cli_subcommand,
+                long_option_with_underscores,
+                expected_default_value_converted,
+            )
+
+    # Helper methods
+
+    def _test_long_flag(
+        self,
+        long_option,
+        cli_subcommand,
+        expected_value_string,
+        long_option_with_underscores,
+        expected_value,
+        extra_commands=None,
+    ):
+        cli = cli_subcommand()
+        cli.args.extend([long_option, expected_value_string])
+        if extra_commands is not None:
+            cli.args.extend(extra_commands)
+        _, config = cli.parse()
+        option_value = config.get_config().get(long_option_with_underscores).value()
+        self.assertEqual(option_value, expected_value)
+
+    def _test_short_flag(
+        self,
+        short_option,
+        cli_subcommand,
+        expected_value_string,
+        long_option_with_underscores,
+        expected_value,
+    ):
+        if short_option is not None:
+            cli = cli_subcommand()
+            cli.args.extend([short_option, expected_value_string])
+            _, config = cli.parse()
+            option_value = config.get_config().get(long_option_with_underscores).value()
+            self.assertEqual(option_value, expected_value)
+
+    def _test_expected_default_value(
+        self, cli_subcommand, long_option_with_underscores, expected_default_value
+    ):
+        cli = cli_subcommand()
+        _, config = cli.parse()
+        option_value = (
+            config.get_config().get(long_option_with_underscores).default_value()
+        )
+        self.assertEqual(option_value, expected_default_value)
+
+    def _convert_flag_to_use_underscores(self, option):
+        return option.lstrip("-").replace("-", "_")
+
+    def _convert_string_to_numeric(self, number):
+        return float(number) if "." in number else int(number)
+
+    def _convert_string_to_int_list(self, list_values):
+        ret_val = [int(x) for x in list_values.split(",")]
+        if len(ret_val) == 1:
+            return ret_val[0]
+        return ret_val
+
+    def _convert_string_to_string_list(self, list_values):
+        ret_val = [x for x in list_values.split(",")]
+        if len(ret_val) == 1:
+            return ret_val[0]
+        return ret_val
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_config.py b/src/c++/perf_analyzer/tests/test_config.py
new file mode 100755
index 000000000..f056eb76a
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_config.py
@@ -0,0 +1,2382 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import re
+import unittest
+from argparse import Namespace
+from copy import deepcopy
+from typing import Any, Dict, List, Optional
+from unittest.mock import patch
+
+from model_analyzer.cli.cli import CLI
+from model_analyzer.config.input.config_command_profile import ConfigCommandProfile
+from model_analyzer.config.input.config_command_report import ConfigCommandReport
+from model_analyzer.config.input.config_enum import ConfigEnum
+from model_analyzer.config.input.config_list_generic import ConfigListGeneric
+from model_analyzer.config.input.config_list_numeric import ConfigListNumeric
+from model_analyzer.config.input.config_list_string import ConfigListString
+from model_analyzer.config.input.config_object import ConfigObject
+from model_analyzer.config.input.config_primitive import ConfigPrimitive
+from model_analyzer.config.input.config_sweep import ConfigSweep
+from model_analyzer.config.input.config_union import ConfigUnion
+from model_analyzer.config.input.objects.config_model_profile_spec import (
+    ConfigModelProfileSpec,
+)
+from model_analyzer.config.input.objects.config_plot import ConfigPlot
+from model_analyzer.constants import CONFIG_PARSER_FAILURE
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.result.model_constraints import ModelConstraints
+
+from .common import test_result_collector as trc
+from .mocks.mock_config import MockConfig
+from .mocks.mock_numba import MockNumba
+from .mocks.mock_os import MockOSMethods
+
+
+class TestConfig(trc.TestResultCollector):
+    def _create_parameters(
+        self, batch_sizes: List = [], concurrency: List = [], request_rate: List = []
+    ) -> Dict:
+        return {
+            "batch_sizes": batch_sizes,
+            "concurrency": concurrency,
+            "request_rate": request_rate,
+        }
+
+    def _evaluate_config(
+        self, args, yaml_content, subcommand="profile", numba_available=True
+    ):
+        mock_numba = MockNumba(
+            mock_paths=["model_analyzer.config.input.config_command_profile"],
+            is_available=numba_available,
+        )
+
+        mock_config = MockConfig(args, yaml_content)
+        mock_config.start()
+        mock_numba.start()
+
+        if subcommand == "report":
+            config = ConfigCommandReport()
+        else:
+            config = ConfigCommandProfile()
+        cli = CLI()
+        cli.add_subcommand(cmd=subcommand, config=config, help="Test subcommand help")
+        cli.parse()
+        mock_config.stop()
+        mock_numba.stop()
+        return config
+
+    def _assert_error_on_evaluate_config(
+        self, args, yaml_content, subcommand="profile"
+    ):
+        mock_config = MockConfig(args, yaml_content)
+        mock_config.start()
+        if subcommand == "report":
+            config = ConfigCommandReport()
+        else:
+            config = ConfigCommandProfile()
+        cli = CLI()
+        cli.add_subcommand(cmd=subcommand, config=config, help="Test subcommand help")
+        # When a required field is not specified, parse will lead to an
+        # exception
+        with self.assertRaises(TritonModelAnalyzerException):
+            cli.parse()
+        mock_config.stop()
+
+    def _assert_equality_of_model_configs(self, model_configs, expected_model_configs):
+        self.assertEqual(len(model_configs), len(expected_model_configs))
+        for model_config, expected_model_config in zip(
+            model_configs, expected_model_configs
+        ):
+            self.assertEqual(expected_model_config.cpu_only(), model_config.cpu_only())
+            self.assertEqual(
+                expected_model_config.model_name(), model_config.model_name()
+            )
+            self.assertEqual(
+                expected_model_config.parameters(), model_config.parameters()
+            )
+            self.assertEqual(
+                expected_model_config.constraints(), model_config.constraints()
+            )
+            self.assertEqual(
+                expected_model_config.objectives(), model_config.objectives()
+            )
+            self.assertEqual(
+                expected_model_config.model_config_parameters(),
+                model_config.model_config_parameters(),
+            )
+
+    def _assert_equality_of_plot_configs(self, plot_configs, expected_plot_configs):
+        self.assertEqual(len(plot_configs), len(expected_plot_configs))
+        for plot_config, expected_plot_config in zip(
+            plot_configs, expected_plot_configs
+        ):
+            self.assertEqual(expected_plot_config.name(), plot_config.name())
+            self.assertEqual(expected_plot_config.title(), plot_config.title())
+            self.assertEqual(expected_plot_config.x_axis(), plot_config.x_axis())
+            self.assertEqual(expected_plot_config.y_axis(), plot_config.y_axis())
+            self.assertEqual(expected_plot_config.monotonic(), plot_config.monotonic())
+
+    def _assert_model_config_types(self, model_config):
+        self.assertIsInstance(model_config.field_type(), ConfigUnion)
+
+        if isinstance(model_config.field_type().raw_value(), ConfigListGeneric):
+            self.assertIsInstance(
+                model_config.field_type().raw_value().container_type(), ConfigUnion
+            )
+        else:
+            self.assertIsInstance(model_config.field_type().raw_value(), ConfigObject)
+
+    def _assert_model_object_types(
+        self,
+        model_config,
+        model_name,
+        check_parameters=False,
+        check_concurrency=False,
+        check_batch_size=False,
+    ):
+        if isinstance(model_config, ConfigUnion):
+            self.assertIsInstance(model_config.raw_value(), ConfigObject)
+            self.assertIsInstance(
+                model_config.raw_value().raw_value()[model_name], ConfigObject
+            )
+        else:
+            self.assertIsInstance(model_config, ConfigObject)
+            if check_parameters:
+                parameters_config = model_config.raw_value()["parameters"]
+                self.assertIsInstance(parameters_config, ConfigObject)
+
+                if check_concurrency:
+                    self.assertIsInstance(
+                        parameters_config.raw_value()["concurrency"], ConfigListNumeric
+                    )
+
+    def _assert_model_config_params(self, model_config_parameters):
+        self.assertIsInstance(model_config_parameters, ConfigObject)
+
+        input_param = model_config_parameters.raw_value()["input"]
+        self.assertIsInstance(input_param, ConfigSweep)
+        # Is list of params
+        self.assertIsInstance(input_param.raw_value(), ConfigListGeneric)
+
+        # Each subitem is also a list
+        self.assertIsInstance(input_param.raw_value().container_type(), ConfigUnion)
+
+        single_sweep_param = input_param.raw_value().raw_value()[0].raw_value()
+        self.assertIsInstance(single_sweep_param.raw_value()[0], ConfigObject)
+
+        # Check types for 'name'
+        name_param = single_sweep_param.raw_value()[0].raw_value()["name"]
+        self.assertIsInstance(name_param, ConfigSweep)
+        self.assertIsInstance(name_param.raw_value(), ConfigListGeneric)
+        self.assertIsInstance(name_param.raw_value().container_type(), ConfigUnion)
+
+        # Check types for 'data_type'
+        data_type_param = single_sweep_param.raw_value()[0].raw_value()["data_type"]
+        self.assertIsInstance(data_type_param, ConfigSweep)
+        self.assertIsInstance(data_type_param.raw_value(), ConfigListGeneric)
+        self.assertIsInstance(data_type_param.raw_value().container_type(), ConfigUnion)
+        self.assertIsInstance(data_type_param.raw_value().raw_value()[0], ConfigUnion)
+        self.assertIsInstance(
+            data_type_param.raw_value().raw_value()[0].raw_value(), ConfigEnum
+        )
+
+        # Check types for 'dims'
+        dims_param = single_sweep_param.raw_value()[0].raw_value()["dims"]
+        self.assertIsInstance(dims_param, ConfigSweep)
+        self.assertIsInstance(dims_param.raw_value(), ConfigListGeneric)
+        self.assertIsInstance(dims_param.raw_value().container_type(), ConfigUnion)
+        self.assertIsInstance(dims_param.raw_value().raw_value()[0], ConfigUnion)
+        self.assertIsInstance(
+            dims_param.raw_value().raw_value()[0].raw_value(), ConfigListGeneric
+        )
+        self.assertIsInstance(
+            dims_param.raw_value().raw_value()[0].raw_value().raw_value()[0],
+            ConfigPrimitive,
+        )
+
+        # Check types for 'format'
+        format_param = single_sweep_param.raw_value()[0].raw_value()["format"]
+        self.assertIsInstance(format_param, ConfigSweep)
+        self.assertIsInstance(format_param.raw_value(), ConfigListGeneric)
+        self.assertIsInstance(format_param.raw_value().container_type(), ConfigUnion)
+
+    def _assert_model_str_type(self, model_config):
+        self.assertIsInstance(model_config, ConfigUnion)
+        self.assertIsInstance(model_config.raw_value(), ConfigPrimitive)
+
+    def setUp(self):
+        # Mock path validation
+        self.mock_os = MockOSMethods(
+            mock_paths=["model_analyzer.config.input.config_utils"]
+        )
+        self.mock_os.start()
+
+    def tearDown(self):
+        self.mock_os.stop()
+        patch.stopall()
+
+    def test_config(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+            "--profile-models",
+            "vgg11",
+        ]
+        yaml_content = "model_repository: yaml_repository"
+        config = self._evaluate_config(args, yaml_content)
+
+        # CLI flag has the highest priority
+        self.assertEqual(config.get_all_config()["model_repository"], "cli_repository")
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "-f",
+            "path-to-config-file",
+            "--profile-models",
+            "vgg11",
+        ]
+        yaml_content = "model_repository: yaml_repository"
+        config = self._evaluate_config(args, yaml_content)
+
+        # If CLI flag doesn't exist, YAML config has the highest priority
+        self.assertEqual(config.get_all_config()["model_repository"], "yaml_repository")
+
+        args = ["model-analyzer", "profile", "-f", "path-to-config-file"]
+        yaml_content = "model_repository: yaml_repository"
+        self._assert_error_on_evaluate_config(args, yaml_content)
+
+    def test_missing_config(self):
+        """Test that we fail if the required option profile-models is omitted"""
+
+        args = ["model-analyzer", "profile", "-f", "path-to-config-file"]
+        yaml_content = "model_repository: yaml_repository"
+        self._assert_error_on_evaluate_config(args, yaml_content)
+
+    def test_conflicting_configs(self):
+        """Test that we fail if an option is specified in both CLI and YAML"""
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+            "--profile-models",
+            "vgg11",
+        ]
+        yaml_content = "profile_models: add_sub"
+        self._assert_error_on_evaluate_config(args, yaml_content)
+
+    def test_range_and_list_values(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = "profile_models: model_1,model_2"
+        config = self._evaluate_config(args, yaml_content)
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "model_1",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+            ),
+            ConfigModelProfileSpec(
+                "model_2",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+            ),
+        ]
+        self._assert_equality_of_model_configs(
+            config.get_all_config()["profile_models"], expected_model_configs
+        )
+        self.assertIsInstance(
+            config.get_config()["profile_models"].field_type().raw_value(), ConfigObject
+        )
+
+        yaml_content = """
+profile_models:
+    - model_1
+    - model_2
+"""
+        config = self._evaluate_config(args, yaml_content)
+        self._assert_equality_of_model_configs(
+            config.get_all_config()["profile_models"], expected_model_configs
+        )
+
+        model_config = config.get_config()["profile_models"]
+
+        self._assert_model_config_types(model_config)
+        self.assertIsInstance(
+            model_config.field_type()
+            .raw_value()
+            .raw_value()["model_1"]
+            .raw_value()["objectives"],
+            ConfigUnion,
+        )
+        self.assertIsInstance(
+            model_config.field_type()
+            .raw_value()
+            .raw_value()["model_1"]
+            .raw_value()["parameters"],
+            ConfigObject,
+        )
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+            "--profile-models",
+            "model_1,model_2",
+        ]
+        yaml_content = """
+batch_sizes:
+    - 2
+    - 3
+"""
+        config = self._evaluate_config(args, yaml_content)
+        self.assertEqual(config.get_all_config()["batch_sizes"], [2, 3])
+        self.assertIsInstance(
+            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
+        )
+
+        yaml_content = """
+batch_sizes: 2
+"""
+        config = self._evaluate_config(args, yaml_content)
+        self.assertEqual(config.get_all_config()["batch_sizes"], [2])
+        self.assertIsInstance(
+            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
+        )
+
+        yaml_content = """
+concurrency: 2
+"""
+        config = self._evaluate_config(args, yaml_content)
+        self.assertEqual(config.get_all_config()["concurrency"], [2])
+        self.assertIsInstance(
+            config.get_config()["concurrency"].field_type(), ConfigListNumeric
+        )
+        self.assertEqual(config.get_all_config()["batch_sizes"], [1])
+        self.assertIsInstance(
+            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
+        )
+
+        yaml_content = """
+batch_sizes:
+    start: 2
+    stop: 6
+"""
+        config = self._evaluate_config(args, yaml_content)
+        self.assertEqual(config.get_all_config()["batch_sizes"], [2, 3, 4, 5, 6])
+        self.assertIsInstance(
+            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
+        )
+
+        yaml_content = """
+batch_sizes:
+    start: 2
+    stop: 6
+    step: 2
+"""
+        config = self._evaluate_config(args, yaml_content)
+        self.assertEqual(config.get_all_config()["batch_sizes"], [2, 4, 6])
+        self.assertIsInstance(
+            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
+        )
+
+    def test_object(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+profile_models:
+  -
+    vgg_16_graphdef:
+      parameters:
+        concurrency:
+          - 1
+          - 2
+          - 3
+          - 4
+  - vgg_19_graphdef
+"""
+        config = self._evaluate_config(args, yaml_content)
+        model_config = config.get_config()["profile_models"]
+        self._assert_model_config_types(model_config)
+
+        expected_model_objects = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={
+                    "batch_sizes": [1],
+                    "concurrency": [1, 2, 3, 4],
+                    "request_rate": [],
+                },
+                objectives={"perf_throughput": 10},
+            ),
+            ConfigModelProfileSpec(
+                "vgg_19_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+            ),
+        ]
+
+        # Check the types for the first value
+        first_model = (
+            model_config.field_type().raw_value().raw_value()["vgg_16_graphdef"]
+        )
+        self._assert_model_object_types(
+            first_model,
+            "vgg_16_graphdef",
+            check_parameters=True,
+            check_concurrency=True,
+        )
+
+        # Check the types for the second value
+        second_model = (
+            model_config.field_type().raw_value().raw_value()["vgg_19_graphdef"]
+        )
+        self._assert_model_object_types(second_model, "vgg_19_graphdef")
+
+        self._assert_equality_of_model_configs(
+            config.get_all_config()["profile_models"], expected_model_objects
+        )
+
+        yaml_content = """
+profile_models:
+  vgg_16_graphdef:
+    parameters:
+      concurrency:
+        - 1
+        - 2
+        - 3
+        - 4
+  vgg_19_graphdef:
+    parameters:
+      concurrency:
+        - 1
+        - 2
+        - 3
+        - 4
+      batch_sizes:
+          start: 2
+          stop: 6
+          step: 2
+"""
+        expected_model_objects = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={
+                    "batch_sizes": [1],
+                    "concurrency": [1, 2, 3, 4],
+                    "request_rate": [],
+                },
+                objectives={"perf_throughput": 10},
+            ),
+            ConfigModelProfileSpec(
+                "vgg_19_graphdef",
+                parameters={
+                    "concurrency": [1, 2, 3, 4],
+                    "batch_sizes": [2, 4, 6],
+                    "request_rate": [],
+                },
+                objectives={"perf_throughput": 10},
+            ),
+        ]
+        config = self._evaluate_config(args, yaml_content)
+        self._assert_equality_of_model_configs(
+            config.get_all_config()["profile_models"], expected_model_objects
+        )
+
+        model_config = config.get_config()["profile_models"]
+        self._assert_model_config_types(model_config)
+
+        # first model
+        first_model = (
+            model_config.field_type().raw_value().raw_value()["vgg_16_graphdef"]
+        )
+        self._assert_model_object_types(
+            first_model,
+            "vgg_16_graphdef",
+            check_parameters=True,
+            check_concurrency=True,
+            check_batch_size=True,
+        )
+
+        # second model
+        second_model = (
+            model_config.field_type().raw_value().raw_value()["vgg_19_graphdef"]
+        )
+        self._assert_model_object_types(
+            second_model,
+            "vgg_19_graphdef",
+            check_parameters=True,
+            check_concurrency=True,
+            check_batch_size=True,
+        )
+
+    def test_constraints(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+profile_models:
+  -
+    vgg_16_graphdef:
+      parameters:
+        concurrency:
+          - 1
+          - 2
+          - 3
+          - 4
+      objectives:
+        perf_throughput: 10
+        gpu_used_memory: 5
+      constraints:
+        gpu_used_memory:
+          max: 80
+  - vgg_19_graphdef
+"""
+        config = self._evaluate_config(args, yaml_content)
+        expected_model_objects = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={
+                    "batch_sizes": [1],
+                    "concurrency": [1, 2, 3, 4],
+                    "request_rate": [],
+                },
+                objectives={"perf_throughput": 10, "gpu_used_memory": 5},
+                constraints={
+                    "gpu_used_memory": {
+                        "max": 80,
+                    }
+                },
+            ),
+            ConfigModelProfileSpec(
+                "vgg_19_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+            ),
+        ]
+        self._assert_equality_of_model_configs(
+            config.get_all_config()["profile_models"], expected_model_objects
+        )
+
+        # GPU Memory shouldn't have min
+        yaml_content = """
+profile_models:
+  -
+    vgg_16_graphdef:
+      parameters:
+        concurrency:
+          - 1
+          - 2
+          - 3
+          - 4
+      objectives:
+        - perf_throughput
+        - gpu_used_memory
+      constraints:
+        gpu_memory:
+          max: 80
+          min: 45
+  - vgg_19_graphdef
+"""
+        self._assert_error_on_evaluate_config(args, yaml_content)
+
+        # Test objective key that is not one of the supported metrics
+        yaml_content = """
+profile_models:
+  -
+    vgg_16_graphdef:
+      parameters:
+        concurrency:
+          - 1
+          - 2
+          - 3
+          - 4
+      objectives:
+        - throughput
+      constraints:
+        gpu_used_memory:
+          max: 80
+  - vgg_19_graphdef
+"""
+        self._assert_error_on_evaluate_config(args, yaml_content)
+
+    def test_validation(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+
+        # end key should not be included in concurrency
+        yaml_content = """
+profile_models:
+  -
+    vgg_16_graphdef:
+      parameters:
+        concurrency:
+            start: 4
+            stop: 12
+            end: 2
+"""
+        self._assert_error_on_evaluate_config(args, yaml_content)
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+
+        yaml_content = """
+profile_models:
+  -
+    vgg_16_graphdef:
+      parameters:
+        concurrency:
+            start: 13
+            stop: 12
+"""
+        self._assert_error_on_evaluate_config(args, yaml_content)
+
+    def test_config_model(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+profile_models:
+  -
+    vgg_16_graphdef:
+        model_config_parameters:
+            instance_group:
+                -
+                    kind: KIND_GPU
+                    count: 1
+
+"""
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+                model_config_parameters={
+                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        yaml_content = """
+profile_models:
+  -
+    vgg_16_graphdef:
+        model_config_parameters:
+            instance_group:
+                -
+                    kind: KIND_GPU
+                    count: 1
+
+"""
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+                model_config_parameters={
+                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+            profile_models:
+            -
+                vgg_16_graphdef:
+                    model_config_parameters:
+                        instance_group:
+                            -
+                                kind: KIND_GPU
+                                count: 1
+                            -
+                                kind: KIND_CPU
+                                count: 1
+
+            """
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+                model_config_parameters={
+                    "instance_group": [
+                        [
+                            {"kind": ["KIND_GPU"], "count": [1]},
+                            {"kind": ["KIND_CPU"], "count": [1]},
+                        ]
+                    ]
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+            profile_models:
+            -
+                vgg_16_graphdef:
+                    model_config_parameters:
+                        instance_group:
+                        -
+                            -
+                                kind: KIND_GPU
+                                count: 1
+                        -
+                            -
+                                kind: KIND_CPU
+                                count: 1
+
+            """
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+                model_config_parameters={
+                    "instance_group": [
+                        [{"kind": ["KIND_GPU"], "count": [1]}],
+                        [{"kind": ["KIND_CPU"], "count": [1]}],
+                    ]
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        yaml_content = """
+            profile_models:
+            -
+                vgg_16_graphdef:
+                    model_config_parameters:
+                        input:
+                        -
+                            name: NV_MODEL_INPUT
+                            data_type: TYPE_FP32
+                            format: FORMAT_NHWC
+                            dims: [256, 256, 3]
+
+            """
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+                model_config_parameters={
+                    "input": [
+                        [
+                            {
+                                "name": ["NV_MODEL_INPUT"],
+                                "data_type": ["TYPE_FP32"],
+                                "format": ["FORMAT_NHWC"],
+                                "dims": [[256, 256, 3]],
+                            }
+                        ]
+                    ]
+                },
+            )
+        ]
+
+        model_config = config.get_config()["profile_models"]
+        self._assert_model_config_types(model_config)
+
+        model = model_config.field_type().raw_value().raw_value()["vgg_16_graphdef"]
+        self._assert_model_object_types(model, "vgg_16_graphdef")
+
+        model_config_parameters = model.raw_value()["model_config_parameters"]
+
+        self._assert_model_config_params(model_config_parameters)
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        yaml_content = """
+            profile_models:
+            -
+                vgg_16_graphdef:
+                    perf_analyzer_flags:
+                        measurement-interval: 10000
+                        model-version: 2
+                        streaming: "header:value"
+
+            """
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+                perf_analyzer_flags={
+                    "measurement-interval": 10000,
+                    "model-version": 2,
+                    "streaming": "header:value",
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        yaml_content = """
+            profile_models:
+            -
+                vgg_16_graphdef:
+                    perf_analyzer_flags:
+                        measurement-interval: 10000
+                        model-version: 2
+                        shape: ["name1:1,2,3", "name2:4,5,6"]
+
+            """
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+                perf_analyzer_flags={
+                    "measurement-interval": 10000,
+                    "model-version": 2,
+                    "shape": ["name1:1,2,3", "name2:4,5,6"],
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        yaml_content = """
+            profile_models:
+            -
+                vgg_16_graphdef:
+                    perf_analyzer_flags:
+                        latency_report_file: ["file1", "file2"]
+
+            """
+        with self.assertRaises(TritonModelAnalyzerException):
+            # latency_report_file is not additive
+            self._evaluate_config(args, yaml_content)
+
+        yaml_content = """
+            profile_models:
+            -
+                vgg_16_graphdef:
+                    perf_analyzer_flags:
+                        disallowed-perf-flag: some_value
+
+            """
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content)
+
+    def test_config_sweep(self):
+        config_sweep = ConfigSweep(ConfigPrimitive(int))
+        config_sweep.set_value(2)
+
+    def test_config_plot(self):
+        args = ["model-analyzer", "report", "-f", "path-to-config-file"]
+        yaml_content = """
+            report_model_configs: vgg_16_graphdef
+            plots:
+                test_plot:
+                    title: Throughput vs. Latency
+                    x_axis: perf_throughput
+                    y_axis: perf_latency_p99
+            """
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        plot_configs = config.get_all_config()["plots"]
+        expected_plot_configs = [
+            ConfigPlot(
+                "test_plot",
+                title="Throughput vs. Latency",
+                x_axis="perf_throughput",
+                y_axis="perf_latency_p99",
+            )
+        ]
+        self._assert_equality_of_plot_configs(plot_configs, expected_plot_configs)
+
+        yaml_content = """
+            report_model_configs: vgg_16_graphdef
+            plots:
+            - test_plot1:
+                title: Throughput vs. Latency
+                x_axis: perf_throughput
+                y_axis: perf_latency_p99
+            - test_plot2:
+                title: GPU Memory vs. Latency
+                x_axis: gpu_used_memory
+                y_axis: perf_latency_p99
+            """
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        plot_configs = config.get_all_config()["plots"]
+        expected_plot_configs = [
+            ConfigPlot(
+                "test_plot1",
+                title="Throughput vs. Latency",
+                x_axis="perf_throughput",
+                y_axis="perf_latency_p99",
+            ),
+            ConfigPlot(
+                "test_plot2",
+                title="GPU Memory vs. Latency",
+                x_axis="gpu_used_memory",
+                y_axis="perf_latency_p99",
+            ),
+        ]
+        self._assert_equality_of_plot_configs(plot_configs, expected_plot_configs)
+
+        yaml_content = """
+            report_model_configs: vgg_16_graphdef
+            plots:
+                test_plot1:
+                    title: Throughput vs. Latency
+                    x_axis: perf_throughput
+                    y_axis: perf_latency_p99
+                test_plot2:
+                    title: GPU Memory vs. Latency
+                    x_axis: gpu_used_memory
+                    y_axis: perf_latency_p99
+            """
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        plot_configs = config.get_all_config()["plots"]
+        expected_plot_configs = [
+            ConfigPlot(
+                "test_plot1",
+                title="Throughput vs. Latency",
+                x_axis="perf_throughput",
+                y_axis="perf_latency_p99",
+            ),
+            ConfigPlot(
+                "test_plot2",
+                title="GPU Memory vs. Latency",
+                x_axis="gpu_used_memory",
+                y_axis="perf_latency_p99",
+            ),
+        ]
+        self._assert_equality_of_plot_configs(plot_configs, expected_plot_configs)
+
+        yaml_content = """
+            report_model_configs:
+              vgg_16_graphdef:
+                plots:
+                    test_plot1:
+                        title: Throughput vs. Latency
+                        x_axis: perf_throughput
+                        y_axis: perf_latency_p99
+                        monotonic: True
+                    test_plot2:
+                        title: GPU Memory vs. Latency
+                        x_axis: gpu_used_memory
+                        y_axis: perf_latency_p99
+                        monotonic: False
+            """
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        plot_configs = config.get_all_config()["report_model_configs"][0].plots()
+        expected_plot_configs = [
+            ConfigPlot(
+                "test_plot1",
+                title="Throughput vs. Latency",
+                x_axis="perf_throughput",
+                y_axis="perf_latency_p99",
+                monotonic=True,
+            ),
+            ConfigPlot(
+                "test_plot2",
+                title="GPU Memory vs. Latency",
+                x_axis="gpu_used_memory",
+                y_axis="perf_latency_p99",
+                monotonic=False,
+            ),
+        ]
+        self._assert_equality_of_plot_configs(plot_configs, expected_plot_configs)
+
+    def test_error_messages(self):
+        # ConfigListNumeric
+        config_numeric = ConfigListNumeric(float)
+        config_numeric.set_name("key")
+        config_status = config_numeric.set_value(
+            {"start": 12, "stop": 15, "undefined_key": 8}
+        )
+
+        config_message = config_status.message()
+        result = re.search(".*'start'.*'stop'.*'undefined_key'.*", config_message)
+
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+        self.assertIsNotNone(config_message)
+        self.assertIsNotNone(result)
+
+        config_numeric = ConfigListNumeric(float)
+        config_numeric.set_name("key")
+        config_status = config_numeric.set_value({"start": 12, "stop": "two"})
+        print(config_message)
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        config_numeric = ConfigListNumeric(float)
+        config_numeric.set_name("key")
+        config_status = config_numeric.set_value({"start": "five", "stop": 2})
+        config_message = config_status.message()
+        print(config_message)
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        config_numeric = ConfigListNumeric(float)
+        config_numeric.set_name("key")
+        config_status = config_numeric.set_value({"start": 10, "stop": 2})
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        # ConfigUnion error message
+        config_union = ConfigUnion([ConfigListNumeric(float), ConfigPrimitive(str)])
+        config_union.set_name("key")
+
+        # Dictionaries are not accepted.
+        config_status = config_union.set_value({"a": "b"})
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        # ConfigEnum
+        config_enum = ConfigEnum(["a", "b"])
+        config_enum.set_name("key")
+        config_status = config_enum.set_value("c")
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        # ConfigListGeneric
+        config_list_generic = ConfigListGeneric(ConfigPrimitive(float))
+        config_list_generic.set_name("key")
+        config_status = config_list_generic.set_value({"a": "b"})
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        # ConfigListString
+        config_list_string = ConfigListString()
+        config_list_string.set_name("key")
+        config_status = config_list_string.set_value({"a": "b"})
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        config_status = config_list_string.set_value([{"a": "b"}])
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        # ConfigObject
+        config_object = ConfigObject(schema={"key": ConfigPrimitive(float)})
+        config_object.set_name("key")
+        config_status = config_object.set_value({"undefiend_key": 2.0})
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        config_status = config_object.set_value({"key": [1, 2, 3]})
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        config_status = config_object.set_value([1, 2, 3])
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+        # ConfigPrimitive
+        config_primitive = ConfigPrimitive(float)
+        config_primitive.set_name("key")
+        config_status = config_primitive.set_value("a")
+        print(config_status.message())
+        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
+
+    def test_autofill(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+        profile_models:
+        -
+            vgg_16_graphdef:
+                model_config_parameters:
+                    instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: 1
+
+        """
+        # Test defaults
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
+                objectives={"perf_throughput": 10},
+                model_config_parameters={
+                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        yaml_content = """
+objectives:
+  perf_throughput: 10
+  gpu_used_memory: 5
+constraints:
+  gpu_used_memory:
+    max: 80
+profile_models:
+  -
+    vgg_16_graphdef:
+        parameters:
+          batch_sizes:
+            - 16
+            - 32
+          concurrency:
+            start: 2
+            stop: 4
+            step: 2
+        model_config_parameters:
+            instance_group:
+                -
+                    kind: KIND_GPU
+                    count: 1
+
+"""
+        # Test autofill objectives and constraints
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={
+                    "batch_sizes": [16, 32],
+                    "concurrency": [2, 4],
+                    "request_rate": [],
+                },
+                objectives={"perf_throughput": 10, "gpu_used_memory": 5},
+                constraints={
+                    "gpu_used_memory": {
+                        "max": 80,
+                    }
+                },
+                model_config_parameters={
+                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        yaml_content = """
+concurrency:
+  start: 2
+  stop : 4
+  step: 2
+profile_models:
+  -
+    vgg_16_graphdef:
+        parameters:
+          batch_sizes:
+            - 16
+            - 32
+        objectives:
+          gpu_used_memory: 10
+        constraints:
+          perf_latency_p99:
+            max: 8000
+        model_config_parameters:
+            instance_group:
+                -
+                    kind: KIND_GPU
+                    count: 1
+
+"""
+        # Test autofill concurrency
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={
+                    "batch_sizes": [16, 32],
+                    "concurrency": [2, 4],
+                    "request_rate": [],
+                },
+                objectives={"gpu_used_memory": 10},
+                constraints={"perf_latency_p99": {"max": 8000}},
+                model_config_parameters={
+                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        yaml_content = """
+batch_sizes:
+  - 16
+  - 32
+profile_models:
+  -
+    vgg_16_graphdef:
+        parameters:
+          concurrency:
+            start: 2
+            stop : 4
+            step: 2
+        objectives:
+          - gpu_used_memory
+        constraints:
+          perf_latency_p99:
+            max: 8000
+        model_config_parameters:
+            instance_group:
+                -
+                    kind: KIND_GPU
+                    count: 1
+
+"""
+        # Test autofill batch sizes
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={
+                    "batch_sizes": [16, 32],
+                    "concurrency": [2, 4],
+                    "request_rate": [],
+                },
+                objectives={"gpu_used_memory": 10},
+                constraints={"perf_latency_p99": {"max": 8000}},
+                model_config_parameters={
+                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
+                },
+            )
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        yaml_content = """
+objectives:
+  perf_throughput: 10
+  perf_latency_p99: 5
+constraints:
+    perf_latency_p99:
+      max: 8000
+    gpu_used_memory:
+      max: 10000
+batch_sizes:
+  - 16
+  - 32
+concurrency:
+  start: 2
+  stop: 4
+  step: 2
+profile_models:
+  -
+    vgg_16_graphdef:
+        parameters:
+          concurrency:
+            start: 5
+            stop : 7
+        objectives:
+          - gpu_used_memory
+  -
+    vgg_19_graphdef:
+        parameters:
+          batch_sizes:
+            - 1
+            - 2
+        constraints:
+          perf_latency_p99:
+            max: 8000
+"""
+        # Test autofill batch sizes
+        config = self._evaluate_config(args, yaml_content)
+        model_configs = config.get_all_config()["profile_models"]
+        expected_model_configs = [
+            ConfigModelProfileSpec(
+                "vgg_16_graphdef",
+                parameters={
+                    "batch_sizes": [16, 32],
+                    "concurrency": [5, 6, 7],
+                    "request_rate": [],
+                },
+                objectives={"gpu_used_memory": 10},
+                constraints={
+                    "perf_latency_p99": {"max": 8000},
+                    "gpu_used_memory": {"max": 10000},
+                },
+            ),
+            ConfigModelProfileSpec(
+                "vgg_19_graphdef",
+                parameters={
+                    "batch_sizes": [1, 2],
+                    "concurrency": [2, 4],
+                    "request_rate": [],
+                },
+                objectives={"perf_throughput": 10, "perf_latency_p99": 5},
+                constraints={"perf_latency_p99": {"max": 8000}},
+            ),
+        ]
+        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
+
+        # Test autofill CPU_ONLY. It will only be false if no local gpus are available AND we are not in remote mode
+        yaml_content = """
+profile_models:
+  - vgg_16_graphdef
+"""
+        for launch_mode in ["remote", "c_api", "docker", "local"]:
+            for local_gpus_available in [True, False]:
+                new_args = args.copy()
+                new_args.extend(["--triton-launch-mode", launch_mode])
+                config = self._evaluate_config(
+                    new_args, yaml_content, numba_available=local_gpus_available
+                )
+                model_configs = config.get_all_config()["profile_models"]
+                expected_cpu_only = not local_gpus_available and launch_mode != "remote"
+                expected_model_configs = [
+                    ConfigModelProfileSpec(
+                        "vgg_16_graphdef",
+                        cpu_only=expected_cpu_only,
+                        parameters=self._create_parameters(batch_sizes=[1]),
+                        objectives={"perf_throughput": 10},
+                    )
+                ]
+                self._assert_equality_of_model_configs(
+                    model_configs, expected_model_configs
+                )
+
+    def test_config_shorthands(self):
+        """
+        test flags like --latency-budget
+        """
+
+        for constraint_shorthand in [
+            ("--latency-budget", "max", "perf_latency_p99"),
+            ("--min-throughput", "min", "perf_throughput"),
+        ]:
+            args = [
+                "model-analyzer",
+                "profile",
+                "--profile-models",
+                "test_model",
+                constraint_shorthand[0],
+                "40",
+                "--model-repository",
+                "cli_repository",
+            ]
+            # check that global and model specific constraints are filled
+            yaml_content = ""
+            config = self._evaluate_config(args, yaml_content, subcommand="profile")
+            self.assertDictEqual(
+                config.get_all_config()["constraints"],
+                {constraint_shorthand[2]: {constraint_shorthand[1]: 40}},
+            )
+
+            self.assertEqual(
+                config.get_all_config()["profile_models"][0].constraints(),
+                ModelConstraints(
+                    {constraint_shorthand[2]: {constraint_shorthand[1]: 40}}
+                ),
+            )
+
+            # check that model specific constraints are appended to
+            args = [
+                "model-analyzer",
+                "profile",
+                constraint_shorthand[0],
+                "40",
+                "-f",
+                "path-to-config-file",
+                "--model-repository",
+                "cli_repository",
+            ]
+            yaml_content = """
+            profile_models:
+                test_model:
+                    constraints:
+                        gpu_used_memory:
+                            max : 100
+            """
+            config = self._evaluate_config(args, yaml_content, subcommand="profile")
+            self.assertDictEqual(
+                config.get_all_config()["constraints"],
+                {constraint_shorthand[2]: {constraint_shorthand[1]: 40}},
+            )
+            self.assertEqual(
+                config.get_all_config()["profile_models"][0].constraints(),
+                ModelConstraints(
+                    {
+                        constraint_shorthand[2]: {constraint_shorthand[1]: 40},
+                        "gpu_used_memory": {"max": 100},
+                    }
+                ),
+            )
+
+            # check that model specific constraints are replaced
+            yaml_content = f"""
+            profile_models:
+                test_model:
+                    constraints:
+                        {constraint_shorthand[2]}:
+                            {constraint_shorthand[1]} : 100
+            """
+            config = self._evaluate_config(args, yaml_content, subcommand="profile")
+            self.assertEqual(
+                config.get_all_config()["profile_models"][0].constraints(),
+                ModelConstraints(
+                    {constraint_shorthand[2]: {constraint_shorthand[1]: 40}}
+                ),
+            )
+
+            # check that global constraints are appended to
+            yaml_content = """
+            profile_models: test_model
+            constraints:
+                gpu_used_memory:
+                    max : 100
+            """
+
+            config = self._evaluate_config(args, yaml_content, subcommand="profile")
+            self.assertDictEqual(
+                config.get_all_config()["constraints"],
+                {
+                    constraint_shorthand[2]: {constraint_shorthand[1]: 40},
+                    "gpu_used_memory": {"max": 100},
+                },
+            )
+
+            # check that global constraints are replaced
+            yaml_content = f"""
+            profile_models: test_model
+            constraints:
+                {constraint_shorthand[2]}:
+                    {constraint_shorthand[1]} : 100
+            """
+            config = self._evaluate_config(args, yaml_content, subcommand="profile")
+            self.assertDictEqual(
+                config.get_all_config()["constraints"],
+                {constraint_shorthand[2]: {constraint_shorthand[1]: 40}},
+            )
+
+    def test_triton_server_flags(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+            profile_models: model1, model2
+            triton_server_flags:
+                strict-model-config: false
+                backend-config: test_backend_config
+            """
+        config = self._evaluate_config(args, yaml_content)
+        self.assertDictEqual(
+            config.get_all_config()["triton_server_flags"],
+            {"strict-model-config": "False", "backend-config": "test_backend_config"},
+        )
+
+        yaml_content = """
+            profile_models: model1, model2
+            triton_server_flags:
+                disallowed-config-option: some_value
+                backend-config: test_backend_config
+            """
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content)
+
+        yaml_content = """
+            profile_models:
+                model1:
+                    triton_server_flags:
+                        strict_model_config: false
+                        backend_config: test_backend_config
+            """
+        config = self._evaluate_config(args, yaml_content)
+        self.assertDictEqual(
+            config.get_all_config()["profile_models"][0].triton_server_flags(),
+            {"strict_model_config": "False", "backend_config": "test_backend_config"},
+        )
+
+    def test_triton_server_environment(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+            profile_models: model1, model2
+            triton_server_environment:
+                LD_PRELOAD: libtest.so
+                LD_LIBRARY_PATH: /path/to/test/lib
+            """
+        config = self._evaluate_config(args, yaml_content)
+        self.assertDictEqual(
+            config.get_all_config()["triton_server_environment"],
+            {"LD_PRELOAD": "libtest.so", "LD_LIBRARY_PATH": "/path/to/test/lib"},
+        )
+
+        yaml_content = """
+            profile_models:
+                model1:
+                    triton_server_environment:
+                        LD_PRELOAD: libtest.so
+                        LD_LIBRARY_PATH: /path/to/test/lib
+            """
+        config = self._evaluate_config(args, yaml_content)
+        self.assertDictEqual(
+            config.get_all_config()["profile_models"][0].triton_server_environment(),
+            {"LD_PRELOAD": "libtest.so", "LD_LIBRARY_PATH": "/path/to/test/lib"},
+        )
+
+    def test_triton_docker_args(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+            "--triton-launch-mode",
+            "docker",
+        ]
+        yaml_content = """
+            profile_models: model1, model2
+            triton_docker_args:
+                image: foo
+                command: goo
+            """
+        config = self._evaluate_config(args, yaml_content)
+        self.assertDictEqual(
+            config.get_all_config()["triton_docker_args"],
+            {"image": "foo", "command": "goo"},
+        )
+
+        yaml_content = """
+            profile_models:
+                model1:
+                    triton_docker_args:
+                        image: foo
+                        command: goo
+            """
+        config = self._evaluate_config(args, yaml_content)
+        self.assertDictEqual(
+            config.get_all_config()["profile_models"][0].triton_docker_args(),
+            {"image": "foo", "command": "goo"},
+        )
+
+        # This is testing that all options can be parsed by the YAML
+        yaml_content = """
+            profile_models: model1, model2
+            triton_docker_args:
+                image: abc
+                command: abc
+                auto_remove: true
+                blkio_weight_device: ['abc']
+                blkio_weight: 1
+                cap_add: ['abc']
+                cap_drop: ['abc']
+                cgroup_parent: abc
+                cgroupns: abc
+                cpu_count: 1
+                cpu_percent: 1
+                cpu_period: 1
+                cpu_quota: 1
+                cpu_rt_period: 1
+                cpu_shares: 1
+                cpuset_cpus: abc
+                cpuset_mems: abc
+                detach: true
+                domainname: abc
+                entrypoint: abc
+                environment: ['abc']
+                hostname: abc
+                init: true
+                init_path: abc
+                ipc_mode: abc
+                isolation: abc
+                kernel_memory: abc
+                labels: abc
+                mac_address: abc
+                mem_limit: abc
+                mem_reservation: abc
+                memswap_limit: abc
+                name: abc
+                nano_cpus: 1
+                network: abc
+                network_disabled: true
+                network_mode: abc
+                oom_kill_disable: true
+                oom_score_adj: 1
+                pid_mode: abc
+                pids_limit: 1
+                platform: abc
+                privileged: true
+                publish_all_ports: true
+                remove: true
+                runtime: abc
+                shm_size: abc
+                stdin_open: true
+                stdout: true
+                stderr: true
+                stop_signal: abc
+                stream: true
+                tty: true
+                use_config_proxy: true
+                user: abc
+                userns_mode: abc
+                uts_mode: abc
+                version: abc
+                volume_driver: abc
+                volumes: ['abc']
+                working_dir: abc
+            """
+        _ = self._evaluate_config(args, yaml_content)
+
+    def test_report_configs(self):
+        args = ["model-analyzer", "report", "--report-model-configs", "test-model"]
+        yaml_content = ""
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        self.assertEqual(
+            config.get_all_config()["report_model_configs"][0].model_config_name(),
+            "test-model",
+        )
+
+        # check output format
+        args = [
+            "model-analyzer",
+            "report",
+            "--report-model-configs",
+            "test-model",
+            "--output-formats",
+            "pdf",
+        ]
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        self.assertEqual(config.get_all_config()["output_formats"], ["pdf"])
+        args = [
+            "model-analyzer",
+            "report",
+            "--report-model-configs",
+            "test-model",
+            "--output-formats",
+            "pdf,csv,svg",
+        ]
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        self.assertEqual(
+            config.get_all_config()["output_formats"], ["pdf", "csv", "svg"]
+        )
+
+        # Check yaml report model config
+        args = ["model-analyzer", "report", "-f", "path-to-config-file"]
+        yaml_content = """
+        report_model_configs:
+            - test_model_config_0
+        output_formats:
+          - pdf
+          - csv
+          - png
+        """
+
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        self.assertEqual(
+            config.get_all_config()["report_model_configs"][0].model_config_name(),
+            "test_model_config_0",
+        )
+        self.assertEqual(
+            config.get_all_config()["output_formats"], ["pdf", "csv", "png"]
+        )
+
+        # Check plots
+        args = ["model-analyzer", "report", "-f", "path-to-config-file"]
+        yaml_content = """
+        report_model_configs:
+           - test_model_config_0
+           - test_model_config_1
+        plots:
+            throughput_v_latency:
+                title: Throughput vs. Latency
+                x_axis: perf_latency_p99
+                y_axis: perf_throughput
+                monotonic: True
+        """
+
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        self.assertEqual(
+            config.get_all_config()["report_model_configs"][0].model_config_name(),
+            "test_model_config_0",
+        )
+        self.assertEqual(
+            config.get_all_config()["report_model_configs"][1].model_config_name(),
+            "test_model_config_1",
+        )
+        expected_config_plot = {
+            "throughput_v_latency": {
+                "title": "Throughput vs. Latency",
+                "x_axis": "perf_latency_p99",
+                "y_axis": "perf_throughput",
+                "monotonic": True,
+            }
+        }
+        config_plot = config.get_all_config()["plots"][0]
+        config_plot_dict = {
+            config_plot.name(): {
+                "title": config_plot.title(),
+                "x_axis": config_plot.x_axis(),
+                "y_axis": config_plot.y_axis(),
+                "monotonic": config_plot.monotonic(),
+            }
+        }
+        self.assertDictEqual(config_plot_dict, expected_config_plot)
+        for report_model_config in config.report_model_configs:
+            config_plot = report_model_config.plots()[0]
+            config_plot_dict = {
+                config_plot.name(): {
+                    "title": config_plot.title(),
+                    "x_axis": config_plot.x_axis(),
+                    "y_axis": config_plot.y_axis(),
+                    "monotonic": config_plot.monotonic(),
+                }
+            }
+            self.assertDictEqual(config_plot_dict, expected_config_plot)
+
+        # Check individual plots
+        yaml_content = """
+        report_model_configs:
+            test_model_config_0:
+                plots:
+                  model_specific_throughput_v_latency:
+                    title: model specific title
+                    x_axis: perf_latency_p99
+                    y_axis: perf_throughput
+                    monotonic: True
+        plots:
+            throughput_v_latency:
+                title: Throughput vs. Latency
+                x_axis: perf_latency_p99
+                y_axis: perf_throughput
+                monotonic: True
+        """
+
+        config = self._evaluate_config(args, yaml_content, subcommand="report")
+        global_config_plot = config.get_all_config()["plots"][0]
+        global_config_plot_dict = {
+            config_plot.name(): {
+                "title": global_config_plot.title(),
+                "x_axis": global_config_plot.x_axis(),
+                "y_axis": global_config_plot.y_axis(),
+                "monotonic": global_config_plot.monotonic(),
+            }
+        }
+        self.assertDictEqual(expected_config_plot, global_config_plot_dict)
+        model_specific_plot = config.get_all_config()["report_model_configs"][
+            0
+        ].plots()[0]
+        self.assertEqual(
+            model_specific_plot.name(), "model_specific_throughput_v_latency"
+        )
+        self.assertEqual(model_specific_plot.title(), "model specific title")
+
+    def test_path_validation(self):
+        # Test parent path validator
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "/",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+        checkpoint_directory: /test
+        profile_models:
+            - model1
+            - model2
+        """
+
+        self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        self.mock_os.set_os_path_exists_return_value(False)
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+        self.mock_os.set_os_path_exists_return_value(True)
+
+        # Test file path validator
+        yaml_content = """
+        triton_install_path: /opt/triton-model-analyzer/tests/test_config.py
+        profile_models:
+            - model1
+            - model2
+        """
+
+        self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        self.mock_os.set_os_path_exists_return_value(False)
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+        self.mock_os.set_os_path_exists_return_value(True)
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "-f",
+            "path-to-config-file",
+            "--model-repository",
+            "cli_repository",
+        ]
+        yaml_content = """
+        export_path: /opt/triton-model-analyzer/tests
+        profile_models:
+            - model1
+            - model2
+        """
+
+        self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        self.mock_os.set_os_path_exists_return_value(False)
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+        self.mock_os.set_os_path_exists_return_value(True)
+
+        # Test the binary path validator
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "/",
+            "-f",
+            "path-to-config-file",
+        ]
+        yaml_content = """
+        triton_server_path: tritonserver
+        profile_models:
+            - model1
+            - model2
+        """
+
+        self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        self.mock_os.set_os_path_exists_return_value(False)
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+        self.mock_os.set_os_path_exists_return_value(True)
+
+    def test_copy(self):
+        """
+        Test that deepcopy works correctly
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+            "--profile-models",
+            "vgg11",
+        ]
+        yaml_content = "model_repository: yaml_repository"
+        configA = self._evaluate_config(args, yaml_content)
+
+        configB = deepcopy(configA)
+
+        self._assert_equality_of_model_configs(
+            configA.get_all_config()["profile_models"],
+            configB.get_all_config()["profile_models"],
+        )
+
+        self.assertEqual(configA.run_config_search_mode, configB.run_config_search_mode)
+
+        configB.run_config_search_mode = "quick"
+
+        self.assertNotEqual(
+            configA.run_config_search_mode, configB.run_config_search_mode
+        )
+
+    def test_multi_model_search_mode(self):
+        """
+        Test that multi-model is only run in quick
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--profile-models",
+            "test_modelA,test_modelB",
+            "--run-config-profile-models-concurrently-enable",
+        ]
+
+        yaml_content = ""
+
+        # Tests the case where no search mode is specified (default is brute)
+        self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        # Brute should fail
+        new_args = list(args)
+        new_args.append("--run-config-search-mode")
+        new_args.append("brute")
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(new_args, yaml_content, subcommand="profile")
+
+        # Quick should pass
+        new_args = list(args)
+        new_args.append("--run-config-search-mode")
+        new_args.append("quick")
+
+        self._evaluate_config(new_args, yaml_content, subcommand="profile")
+
+    def test_quick_search_mode(self):
+        """
+        Test that only legal options are specified in quick search
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--profile-models",
+            "test_modelA",
+            "--run-config-search-mode",
+            "quick",
+        ]
+
+        yaml_content = ""
+
+        self._evaluate_config(args, yaml_content)
+
+        self._test_quick_search_with_rcs(
+            args,
+            yaml_content,
+            "--run-config-search-disable",
+            use_value=False,
+            use_list=False,
+        )
+        self._test_quick_search_with_rcs(
+            args, yaml_content, "--batch-sizes", use_value=False
+        )
+        self._test_quick_search_with_rcs(
+            args, yaml_content, "--concurrency", use_value=False
+        )
+
+    def test_quick_search_model_specific(self):
+        """
+        Test for illegal model specific options in the YAML during quick search
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--run-config-search-mode",
+            "quick",
+            "-f",
+            "path-to-config-file",
+        ]
+
+        yaml_content = """
+        profile_models:
+          model_1:
+            parameters:
+              concurrency: 1,2,4,128
+        """
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        yaml_content = """
+        profile_models:
+          model_1:
+            parameters:
+              batch size: 1,2,4,128
+        """
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        yaml_content = """
+        profile_models:
+          model_1:
+            model_config_parameters:
+              max_batch_size: 2
+        """
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        yaml_content = """
+        profile_models:
+          model_1:
+            model_config_parameters:
+              instance_group:
+                - kind: KIND_GPU
+        """
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        yaml_content = """
+        profile_models:
+          model_1:
+            model_config_parameters:
+               optimization:
+                 execution_accelerators:
+                   gpu_execution_accelerator:
+                     - name: tensorrt
+                       parameters:
+                       precision_mode: [FP16, FP32]
+        """
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+
+    def test_model_weightings_profile(self):
+        """
+        Test that weightings can be specified only per model in the YAML
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--run-config-search-mode",
+            "quick",
+            "-f",
+            "path-to-config-file",
+        ]
+
+        # Global weighting - illegal
+        yaml_content = """
+        weighting: 3
+
+        profile_models:
+          resnet50_libtorch:
+            objectives:
+              perf_latency_p99: 1
+          add_sub:
+            objectives:
+              perf_throughput: 1
+        """
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+
+        # Model specific weighting - legal
+        yaml_content = """
+        profile_models:
+          resnet50_libtorch:
+            weighting: 3
+            objectives:
+              perf_latency_p99: 1
+          add_sub:
+            weighting: 1
+            objectives:
+              perf_throughput: 1
+        """
+
+        self._evaluate_config(args, yaml_content, subcommand="profile")
+
+    def _test_quick_search_with_rcs(
+        self,
+        args: Namespace,
+        yaml_content: Optional[Dict[str, List]],
+        rcs_string: str,
+        use_value: bool = True,
+        use_list: bool = True,
+    ) -> None:
+        """
+        Tests that run-config-search options raise exceptions
+        in quick search mode
+        """
+        new_args = list(args)
+        new_args.append(rcs_string)
+        if use_value:
+            new_args.append("1")
+        elif use_list:
+            new_args.append(["1", "2", "4"])
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(new_args, yaml_content, subcommand="profile")
+
+    def test_bls_composing_models(self):
+        """
+        Test that BLS composing models can be specified
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--run-config-search-mode",
+            "quick",
+            "--profile-models",
+            "modelA",
+            "--bls-composing-models",
+            "modelA,modelB",
+        ]
+        yaml_content = ""
+
+        self._evaluate_config(args, yaml_content)
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--run-config-search-mode",
+            "quick",
+            "--profile-models",
+            "modelA",
+        ]
+
+        yaml_content = """
+        bls_composing_models: modelA,modelB
+        """
+
+        self._evaluate_config(args, yaml_content)
+
+    def test_bls_illegal_config_combinations(self):
+        """
+        Test BLS illegal config combinations
+        """
+        # BLS with brute search mode
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--run-config-search-mode",
+            "brute",
+            "--profile-models",
+            "modelA",
+            "--bls-composing-models",
+            "modelA,modelB",
+        ]
+        yaml_content = ""
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content)
+
+        # BLS with multiple models
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--run-config-search-mode",
+            "quick",
+            "--profile-models",
+            "modelA,modelB",
+            "--bls-composing-models",
+            "modelA,modelB",
+        ]
+        yaml_content = ""
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content)
+
+        # BLS with concurrent search mode
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--run-config-search-mode",
+            "quick",
+            "--profile-models",
+            "modelA",
+            "--bls-composing-models",
+            "modelA,modelB",
+            "--run-config-profile-models-concurrently-enable",
+        ]
+        yaml_content = ""
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content)
+
+    def test_concurrency_rate_request_config_combinations(self):
+        """
+        Test for concurrency with rate request conflicts
+        """
+        base_args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--profile-models",
+            "modelA",
+            "-c",
+            "1,2,3",
+        ]
+        yaml_content = ""
+
+        self._test_request_rate_config_conflicts(base_args, yaml_content)
+
+    def test_config_search_min_rate_request_config_combinations(self):
+        """
+        Test for concurrency min request with rate request conflicts
+        """
+        base_args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--profile-models",
+            "modelA",
+            "--run-config-search-min-concurrency",
+            "1",
+        ]
+        yaml_content = ""
+
+        self._test_request_rate_config_conflicts(base_args, yaml_content)
+
+    def test_config_search_max_rate_request_config_combinations(self):
+        """
+        Test for concurrency max request with rate request conflicts
+        """
+        base_args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "--profile-models",
+            "modelA",
+            "--run-config-search-max-concurrency",
+            "1",
+        ]
+        yaml_content = ""
+
+        self._test_request_rate_config_conflicts(base_args, yaml_content)
+
+    def test_mixing_request_rate_and_concurrency(self):
+        """
+        Test that mixing models parameters of request-rate and concurrency is illegal
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli-repository",
+            "-f",
+            "path-to-config-file",
+        ]
+
+        yaml_content = """
+        profile_models:
+
+          resnet50_libtorch:
+            parameters:
+              concurrency: 4
+          add_sub:
+            parameters:
+              request_rate: 16
+        """
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content, subcommand="profile")
+
+    def _test_request_rate_config_conflicts(
+        self, base_args: List[Any], yaml_content: str
+    ) -> None:
+        self._test_arg_conflict(
+            base_args, yaml_content, ["--request-rate-search-enable"]
+        )
+        self._test_arg_conflict(base_args, yaml_content, ["--request-rate", "1,2,3"])
+        self._test_arg_conflict(
+            base_args, yaml_content, ["--run-config-search-min-request-rate", "1"]
+        )
+        self._test_arg_conflict(
+            base_args, yaml_content, ["--run-config-search-max-request-rate", "1"]
+        )
+
+    def _test_arg_conflict(
+        self, base_args: List[Any], yaml_content: str, new_args: List[Any]
+    ) -> None:
+        args = base_args.copy() + new_args
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._evaluate_config(args, yaml_content)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_constraint_manager.py b/src/c++/perf_analyzer/tests/test_constraint_manager.py
new file mode 100755
index 000000000..f4e23e7df
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_constraint_manager.py
@@ -0,0 +1,620 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.constants import GLOBAL_CONSTRAINTS_KEY
+from model_analyzer.result.constraint_manager import ConstraintManager
+from model_analyzer.result.model_constraints import ModelConstraints
+
+from .common import test_result_collector as trc
+from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
+
+
+class TestConstraintManager(trc.TestResultCollector):
+    def tearDown(self):
+        patch.stopall()
+
+    def test_single_model_no_constraints(self):
+        """
+        Test that constraints are empty
+        """
+        config = self._create_single_model_no_constraints()
+        constraint_manager = ConstraintManager(config)
+        constraints = constraint_manager.get_constraints_for_all_models()
+
+        self.assertEqual(constraints["model_A"], ModelConstraints({}))
+        self.assertEqual(constraints[GLOBAL_CONSTRAINTS_KEY], ModelConstraints({}))
+
+    def test_single_model_with_constraints(self):
+        """
+        Test that model specific constraints are set
+        """
+        config = self._create_single_model_with_constraints()
+        constraint_manager = ConstraintManager(config)
+        constraints = constraint_manager.get_constraints_for_all_models()
+
+        self.assertEqual(
+            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 100}})
+        )
+        self.assertEqual(constraints[GLOBAL_CONSTRAINTS_KEY], ModelConstraints({}))
+
+    def test_single_model_with_global_constraints(self):
+        """
+        Test that global constraints are attributed to a model
+        """
+        config = self._create_single_model_global_constraints()
+        constraint_manager = ConstraintManager(config)
+        constraints = constraint_manager.get_constraints_for_all_models()
+
+        self.assertEqual(
+            constraints["model_A"], ModelConstraints({"perf_throughput": {"min": 100}})
+        )
+        self.assertEqual(
+            constraints[GLOBAL_CONSTRAINTS_KEY],
+            ModelConstraints({"perf_throughput": {"min": 100}}),
+        )
+
+    def test_single_model_with_both_constraints(self):
+        """
+        Test that model specific constraints override global
+        """
+        config = self._create_single_model_both_constraints()
+        constraint_manager = ConstraintManager(config)
+        constraints = constraint_manager.get_constraints_for_all_models()
+
+        self.assertEqual(
+            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 50}})
+        )
+        self.assertEqual(
+            constraints[GLOBAL_CONSTRAINTS_KEY],
+            ModelConstraints({"perf_latency_p99": {"max": 100}}),
+        )
+
+    def test_multi_model_with_no_global_constraints(self):
+        """
+        Test multi-model with only model constraints and no global constraints
+        """
+        config = self._create_multi_model_with_no_global_constraints()
+        constraint_manager = ConstraintManager(config)
+        constraints = constraint_manager.get_constraints_for_all_models()
+
+        self.assertEqual(
+            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 50}})
+        )
+        self.assertEqual(
+            constraints["model_B"], ModelConstraints({"perf_throughput": {"min": 100}})
+        )
+
+    def test_multi_model_with_matching_global_constraints(self):
+        """
+        Test multi-model with Global constraints and individual overrides
+        """
+        config = self._create_multi_model_with_matching_global_constraints()
+        constraint_manager = ConstraintManager(config)
+        constraints = constraint_manager.get_constraints_for_all_models()
+
+        self.assertEqual(
+            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 75}})
+        )
+        self.assertEqual(
+            constraints["model_B"], ModelConstraints({"perf_throughput": {"min": 125}})
+        )
+        self.assertEqual(
+            constraints[GLOBAL_CONSTRAINTS_KEY],
+            ModelConstraints(
+                {"perf_throughput": {"min": 100}, "gpu_used_memory": {"max": 1000}}
+            ),
+        )
+
+    def test_multi_model_with_different_global_constrants(self):
+        """
+        Test multi-model with different global and individual model constraints
+        """
+        config = self._create_multi_model_with_different_global_constrants()
+        constraint_manager = ConstraintManager(config)
+        constraints = constraint_manager.get_constraints_for_all_models()
+
+        self.assertEqual(
+            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 100}})
+        )
+        self.assertEqual(
+            constraints["model_B"], ModelConstraints({"perf_throughput": {"min": 150}})
+        )
+        self.assertEqual(
+            constraints[GLOBAL_CONSTRAINTS_KEY],
+            ModelConstraints({"gpu_used_memory": {"max": 2000}}),
+        )
+
+    def test_single_model_max_constraint_checks(self):
+        """
+        Test that satisfies_constraints works for a single model
+        with a max style constraint
+        """
+        config = self._create_single_model_with_constraints()
+        constraint_manager = ConstraintManager(config)
+
+        # Constraint is P99 Latency max of 100
+        rcm = self._construct_rcm({"perf_latency_p99": 101}, constraint_manager)
+        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
+
+        rcm = self._construct_rcm({"perf_latency_p99": 100}, constraint_manager)
+        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
+
+        rcm = self._construct_rcm({"perf_latency_p99": 99}, constraint_manager)
+        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
+
+    def test_single_model_min_constraint_checks(self):
+        """
+        Test that satisfies_constraints works for a single model
+        with a min style constraint
+        """
+        config = self._create_single_model_global_constraints()
+        constraint_manager = ConstraintManager(config)
+
+        # Constraint is throughput min of 100
+        rcm = self._construct_rcm({"perf_throughput": 101}, constraint_manager)
+        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
+
+        rcm = self._construct_rcm({"perf_throughput": 100}, constraint_manager)
+        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
+
+        rcm = self._construct_rcm({"perf_throughput": 99}, constraint_manager)
+        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
+
+    def test_multi_model_max_constraint_checks(self):
+        """
+        Test that satisfies_constraints works for multi-model
+        with a max style constraints
+        """
+        # Constraints are:
+        #  Model A: P99 Latency max of 50
+        #  Model B: Throughput min of 100
+        constraint_manager = ConstraintManager(
+            config=self._create_multi_model_with_no_global_constraints()
+        )
+
+        # Model A & B are both at boundaries
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 50, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 100},
+            ],
+            constraint_manager,
+        )
+        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
+
+        # Model A exceeds latency
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 51, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 100},
+            ],
+            constraint_manager,
+        )
+        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
+
+        # Constraints are:
+        #  Model A: P99 Latency max of 75
+        #  Model B: Throughput min of 125
+        constraint_manager = ConstraintManager(
+            config=self._create_multi_model_with_matching_global_constraints()
+        )
+
+        # Model A & B are both at boundaries
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 75, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 125},
+            ],
+            constraint_manager,
+        )
+        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
+
+        # Model A exceeds latency
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 78, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 125},
+            ],
+            constraint_manager,
+        )
+        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
+
+        # Constraints are:
+        #  Model A: P99 Latency max of 100
+        #  Model B: Throughput min of 150
+        constraint_manager = ConstraintManager(
+            config=self._create_multi_model_with_different_global_constrants()
+        )
+
+        # Model A & B are both at boundaries
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 100, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 150},
+            ],
+            constraint_manager,
+        )
+        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
+
+        # Model A exceeds latency
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 105, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 150},
+            ],
+            constraint_manager,
+        )
+        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
+
+    def test_multi_model_min_constraint_checks(self):
+        """
+        Test that satisfies_constraints works for multi-model
+        with a min style constraints
+        """
+        # Constraints are:
+        #  Model A: P99 Latency max of 50
+        #  Model B: Throughput min of 100
+        constraint_manager = ConstraintManager(
+            config=self._create_multi_model_with_no_global_constraints()
+        )
+
+        # Model B doesn't have enough throughput
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 50, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 99},
+            ],
+            constraint_manager,
+        )
+        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
+
+        # Constraints are:
+        #  Model A: P99 Latency max of 75
+        #  Model B: Throughput min of 125
+        constraint_manager = ConstraintManager(
+            config=self._create_multi_model_with_matching_global_constraints()
+        )
+
+        # Model B doesn't have enough throughput
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 75, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 120},
+            ],
+            constraint_manager,
+        )
+        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
+
+        # Constraints are:
+        #  Model A: P99 Latency max of 100
+        #  Model B: Throughput min of 150
+        constraint_manager = ConstraintManager(
+            config=self._create_multi_model_with_different_global_constrants()
+        )
+
+        # Model B doesn't have enough throughput
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 100, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 140},
+            ],
+            constraint_manager,
+        )
+        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
+
+    def test_single_model_max_failure_percentage(self):
+        """
+        Test that constraint_failure_percentage works for a single model
+        with a max style constraint
+        """
+        config = self._create_single_model_with_constraints()
+        constraint_manager = ConstraintManager(config)
+
+        # Constraint is P99 Latency max of 100
+        rcm = self._construct_rcm({"perf_latency_p99": 225}, constraint_manager)
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 125)
+        rcm = self._construct_rcm({"perf_latency_p99": 150}, constraint_manager)
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
+
+        rcm = self._construct_rcm({"perf_latency_p99": 100}, constraint_manager)
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
+
+        rcm = self._construct_rcm({"perf_latency_p99": 99}, constraint_manager)
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
+
+    def test_single_model_min_failure_percentage(self):
+        """
+        Test that constraint_failure_percentage works for a single model
+        with a min style constraint
+        """
+
+        config = self._create_single_model_global_constraints()
+        constraint_manager = ConstraintManager(config)
+
+        # Constraint is throughput min of 100
+        rcm = self._construct_rcm({"perf_throughput": 25}, constraint_manager)
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 75)
+        rcm = self._construct_rcm({"perf_throughput": 50}, constraint_manager)
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
+
+        rcm = self._construct_rcm({"perf_throughput": 100}, constraint_manager)
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
+
+        rcm = self._construct_rcm({"perf_throughput": 101}, constraint_manager)
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
+
+    def test_multi_model_failure_percentage(self):
+        """
+        Test that constraint_failure_percentage works for multi-model
+        with a max style constraints
+        """
+        # Constraints are:
+        #  Model A: P99 Latency max of 50
+        #  Model B: Throughput min of 100
+        constraint_manager = ConstraintManager(
+            config=self._create_multi_model_with_no_global_constraints()
+        )
+
+        # Model A & B are both at boundaries
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 50, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 100},
+            ],
+            constraint_manager,
+        )
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
+
+        # Model A exceeds latency, Model B misses on throughput - each by 20%
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 60, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 80},
+            ],
+            constraint_manager,
+        )
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 40)
+
+        # Model A exceeds latency by 40%, Model B misses on throughput by 10%
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 70, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 90},
+            ],
+            constraint_manager,
+        )
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
+
+        # Constraints are:
+        #  Model A: P99 Latency max of 75
+        #  Model B: Throughput min of 125
+        constraint_manager = ConstraintManager(
+            config=self._create_multi_model_with_matching_global_constraints()
+        )
+
+        # Model A & B are both at boundaries
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 75, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 125},
+            ],
+            constraint_manager,
+        )
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
+
+        # Model A exceeds latency, Model B misses on throughput - each by 25%
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 93.75, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 93.75},
+            ],
+            constraint_manager,
+        )
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
+
+        # Model A exceeds latency by 30%, Model B misses on throughput by 20%
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 97.5, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 100},
+            ],
+            constraint_manager,
+        )
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
+
+        # Constraints are:
+        #  Model A: P99 Latency max of 100
+        #  Model B: Throughput min of 150
+        constraint_manager = ConstraintManager(
+            config=self._create_multi_model_with_different_global_constrants()
+        )
+
+        # Model A & B are both at boundaries
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 100, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 150},
+            ],
+            constraint_manager,
+        )
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
+
+        # Model A exceeds latency, Model B misses on throughput - each by 30%
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 130, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 105},
+            ],
+            constraint_manager,
+        )
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 60)
+
+        # Model A exceeds latency by 35%, Model B misses on throughput by 25%
+        rcm = self._construct_mm_rcm(
+            [
+                {"perf_latency_p99": 135, "perf_throughput": 0},
+                {"perf_latency_p99": 0, "perf_throughput": 112.5},
+            ],
+            constraint_manager,
+        )
+        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 60)
+
+    def _create_single_model_no_constraints(self):
+        args = self._create_args()
+        yaml_str = """
+            profile_models:
+              model_A
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        return config
+
+    def _create_single_model_with_constraints(self):
+        args = self._create_args()
+        yaml_str = """
+            profile_models:
+              model_A:
+                constraints:
+                  perf_latency_p99:
+                    max: 100
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        return config
+
+    def _create_single_model_global_constraints(self):
+        args = self._create_args()
+        yaml_str = """
+            profile_models:
+              model_A
+
+            constraints:
+                perf_throughput:
+                  min: 100
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        return config
+
+    def _create_single_model_both_constraints(self):
+        args = self._create_args()
+        yaml_str = """
+            profile_models:
+              model_A:
+                constraints:
+                  perf_latency_p99:
+                    max: 50
+
+            constraints:
+                perf_latency_p99:
+                  max: 100
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        return config
+
+    def _create_multi_model_with_no_global_constraints(self):
+        args = self._create_args()
+        yaml_str = """
+            profile_models:
+              model_A:
+                constraints:
+                  perf_latency_p99:
+                    max: 50
+              model_B:
+                constraints:
+                  perf_throughput:
+                    min: 100
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        return config
+
+    def _create_multi_model_with_matching_global_constraints(self):
+        args = self._create_args()
+        yaml_str = """
+            profile_models:
+              model_A:
+                constraints:
+                  perf_latency_p99:
+                    max: 75
+              model_B:
+                constraints:
+                  perf_throughput:
+                    min: 125
+            constraints:
+                perf_throughput:
+                  min: 100
+                gpu_used_memory:
+                  max: 1000
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        return config
+
+    def _create_multi_model_with_different_global_constrants(self):
+        args = self._create_args()
+        yaml_str = """
+            profile_models:
+              model_A:
+                constraints:
+                  perf_latency_p99:
+                    max: 100
+              model_B:
+                constraints:
+                  perf_throughput:
+                    min: 150
+            constraints:
+                gpu_used_memory:
+                  max: 2000
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        return config
+
+    def _construct_rcm(self, non_gpu_metric_values, constraint_manager):
+        rcm = construct_run_config_measurement(
+            model_name=MagicMock(),
+            model_config_names=["model_A_config_name_0"],
+            constraint_manager=constraint_manager,
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=MagicMock(),
+            non_gpu_metric_values=[non_gpu_metric_values],
+        )
+
+        return rcm
+
+    def _construct_mm_rcm(self, non_gpu_metric_values, constraint_manager):
+        rcm = construct_run_config_measurement(
+            model_name=MagicMock(),
+            model_config_names=["model_A_config_name_A", "model_B_config_name_B"],
+            constraint_manager=constraint_manager,
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=MagicMock(),
+            non_gpu_metric_values=non_gpu_metric_values,
+        )
+
+        return rcm
+
+    def _create_args(self):
+        return ["model-analyzer", "profile", "-f", "config.yml", "-m", "."]
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_coordinate.py b/src/c++/perf_analyzer/tests/test_coordinate.py
new file mode 100755
index 000000000..156c8bb2f
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_coordinate.py
@@ -0,0 +1,138 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from copy import deepcopy
+
+from model_analyzer.config.generate.coordinate import Coordinate
+
+from .common import test_result_collector as trc
+
+
+class TestCoordinate(trc.TestResultCollector):
+    def test_basic(self):
+        x = Coordinate([0, 0])
+
+        self.assertEqual(x[0], 0)
+        self.assertEqual(x[1], 0)
+
+        x[0] = 2
+        x[1] = 4
+
+        self.assertEqual(x[0], 2)
+        self.assertEqual(x[1], 4)
+
+    def test_out_of_range(self):
+        x = Coordinate([0, 0])
+        with self.assertRaises(IndexError):
+            _ = x[2]
+
+    def test_assignment(self):
+        c1 = Coordinate([2, 4])
+
+        c2 = deepcopy(c1)
+
+        self.assertEqual(c2[0], 2)
+        self.assertEqual(c2[1], 4)
+
+        # Confirm c1 unchanged by c2 update
+        c2[0] = 5
+
+        self.assertEqual(c1[0], 2)
+        self.assertEqual(c1[1], 4)
+
+    def test_coordinate_addition(self):
+        c1 = Coordinate([2, 4])
+        c2 = Coordinate([5, 1])
+
+        c3 = c1 + c2
+        self.assertEqual(c3[0], 7)
+        self.assertEqual(c3[1], 5)
+
+    def test_integer_addition(self):
+        c1 = Coordinate([2, 4])
+        c2 = c1 + 3
+
+        self.assertEqual(c2[0], 5)
+        self.assertEqual(c2[1], 7)
+
+    def test_float_addition(self):
+        c1 = Coordinate([2, 4])
+
+        c2 = c1 + 3.5
+
+        self.assertEqual(c2[0], 5.5)
+        self.assertEqual(c2[1], 7.5)
+
+    def test_division(self):
+        c1 = Coordinate([7, 6])
+
+        c1 /= 2
+
+        self.assertEqual(c1[0], 3.5)
+        self.assertEqual(c1[1], 3)
+
+    def test_multiplication(self):
+        c1 = Coordinate([7, 6])
+
+        c1 *= 2
+
+        self.assertEqual(c1[0], 14)
+        self.assertEqual(c1[1], 12)
+
+    def test_round(self):
+        c1 = Coordinate([0.1, 4.6, 3.9])
+        c1.round()
+        self.assertEqual(c1, Coordinate([0, 5, 4]))
+
+    def test_iterator(self):
+        c1 = Coordinate([2, 4, 1, 7])
+
+        values = []
+        for x in c1:
+            values.append(x)
+
+        self.assertEqual(values, [2, 4, 1, 7])
+
+    def test_enumerate(self):
+        c1 = Coordinate([2, 4, 1, 7])
+
+        indexes = []
+        values = []
+        for i, x in enumerate(c1):
+            indexes.append(i)
+            values.append(x)
+
+        self.assertEqual(indexes, [0, 1, 2, 3])
+        self.assertEqual(values, [2, 4, 1, 7])
+
+    def test_stringification(self):
+        c1 = Coordinate([2, 4, 1, 7])
+        self.assertEqual("[2, 4, 1, 7]", c1.__str__())
+
+    def test_compare(self):
+        c1 = Coordinate([1, 3, 2])
+        c2 = Coordinate([1, 3, 2])
+        c3 = Coordinate([1, 3, 1])
+        c4 = Coordinate([1, 3, 3])
+        c5 = Coordinate([2, 1, 3])
+        c6 = Coordinate([0, 1, 3])
+
+        self.assertEqual(c1, c2)
+        self.assertGreater(c1, c3)
+        self.assertLess(c3, c1)
+        self.assertLess(c1, c4)
+        self.assertLess(c1, c5)
+        self.assertGreater(c1, c6)
diff --git a/src/c++/perf_analyzer/tests/test_coordinate_data.py b/src/c++/perf_analyzer/tests/test_coordinate_data.py
new file mode 100755
index 000000000..a668fd3d1
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_coordinate_data.py
@@ -0,0 +1,123 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock
+
+from model_analyzer.config.generate.coordinate import Coordinate
+from model_analyzer.config.generate.coordinate_data import CoordinateData
+
+from .common import test_result_collector as trc
+from .common.test_utils import construct_run_config_measurement
+
+
+class TestCoordinateData(trc.TestResultCollector):
+    def _construct_rcm(
+        self, throughput: float, latency: float, config_name: str, model_name: str
+    ):
+        model_config_name = [config_name]
+
+        # yapf: disable
+        non_gpu_metric_values = [{
+            "perf_throughput": throughput,
+            "perf_latency_avg": latency
+        }]
+        # yapf: enable
+
+        metric_objectives = [{"perf_throughput": 1}]
+        weights = [1]
+
+        rcm = construct_run_config_measurement(
+            model_name=model_name,
+            model_config_names=model_config_name,
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values={},
+            non_gpu_metric_values=non_gpu_metric_values,
+            metric_objectives=metric_objectives,
+            model_config_weights=weights,
+        )
+        return rcm
+
+    def test_basic(self):
+        result_data = CoordinateData()
+
+        coordinate = Coordinate([0, 0, 0])
+        self.assertEqual(result_data.get_measurement(coordinate), None)
+        self.assertEqual(result_data.get_visit_count(coordinate), 0)
+        self.assertEqual(result_data.is_measured(coordinate), False)
+        self.assertEqual(result_data.has_valid_measurement(coordinate), False)
+
+    def test_visit_count(self):
+        result_data = CoordinateData()
+
+        coordinate1 = Coordinate([0, 0, 0])
+        coordinate2 = Coordinate([0, 4, 1])
+
+        result_data.increment_visit_count(coordinate1)
+        self.assertEqual(1, result_data.get_visit_count(coordinate1))
+
+        result_data.increment_visit_count(coordinate2)
+        self.assertEqual(1, result_data.get_visit_count(coordinate2))
+
+        result_data.increment_visit_count(coordinate1)
+        result_data.increment_visit_count(coordinate1)
+        self.assertEqual(3, result_data.get_visit_count(coordinate1))
+        self.assertEqual(1, result_data.get_visit_count(coordinate2))
+
+    def test_measurement(self):
+        """
+        Test if CoordinateData can properly set and get measurements
+
+        Also confirm that is_measured() and has_valid_measurement() work properly
+        """
+        coordinate_data = CoordinateData()
+
+        coordinate0 = Coordinate([0, 0, 0])
+        coordinate1 = Coordinate([0, 4, 1])
+        coordinate2 = Coordinate([1, 2, 3])
+
+        rcm0 = self._construct_rcm(
+            10, 5, config_name="modelA_config_0", model_name="modelA"
+        )
+        rcm1 = self._construct_rcm(
+            20, 8, config_name="modelB_config_0", model_name="modelB"
+        )
+        rcm2 = None
+
+        coordinate_data.set_measurement(coordinate0, rcm0)
+        coordinate_data.set_measurement(coordinate1, rcm1)
+        coordinate_data.set_measurement(coordinate2, rcm2)
+
+        self.assertEqual(coordinate_data.is_measured(coordinate0), True)
+        self.assertEqual(coordinate_data.is_measured(coordinate1), True)
+        self.assertEqual(coordinate_data.is_measured(coordinate2), True)
+
+        self.assertEqual(coordinate_data.has_valid_measurement(coordinate0), True)
+        self.assertEqual(coordinate_data.has_valid_measurement(coordinate1), True)
+        self.assertEqual(coordinate_data.has_valid_measurement(coordinate2), False)
+
+        measurement0 = coordinate_data.get_measurement(coordinate0)
+        self.assertEqual("modelA_config_0", measurement0.model_variants_name())
+        self.assertEqual(10, measurement0.get_non_gpu_metric_value("perf_throughput"))
+        self.assertEqual(5, measurement0.get_non_gpu_metric_value("perf_latency_avg"))
+        self.assertTrue(coordinate_data.is_measured(coordinate0))
+
+        measurement1 = coordinate_data.get_measurement(coordinate1)
+        self.assertEqual("modelB_config_0", measurement1.model_variants_name())
+        self.assertEqual(20, measurement1.get_non_gpu_metric_value("perf_throughput"))
+        self.assertEqual(8, measurement1.get_non_gpu_metric_value("perf_latency_avg"))
+
+        measurement2 = coordinate_data.get_measurement(coordinate2)
+        self.assertEqual(measurement2, None)
diff --git a/src/c++/perf_analyzer/tests/test_cpu_monitor.py b/src/c++/perf_analyzer/tests/test_cpu_monitor.py
new file mode 100755
index 000000000..030fa9816
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_cpu_monitor.py
@@ -0,0 +1,111 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import time
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.device.gpu_device import GPUDevice
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.monitor.cpu_monitor import CPUMonitor
+from model_analyzer.record.types.cpu_available_ram import CPUAvailableRAM
+from model_analyzer.record.types.cpu_used_ram import CPUUsedRAM
+from model_analyzer.triton.server.server_config import TritonServerConfig
+from model_analyzer.triton.server.server_factory import TritonServerFactory
+
+from .common import test_result_collector as trc
+from .mocks.mock_server_local import MockServerLocalMethods
+
+MODEL_REPOSITORY_PATH = "test_repo"
+TRITON_LOCAL_BIN_PATH = "test_bin_path/tritonserver"
+
+
+class TestCPUMonitor(trc.TestResultCollector):
+    def setUp(self):
+        self.server_local_mock = MockServerLocalMethods()
+        self.server_local_mock.start()
+
+    def test_record_cpu_memory(self):
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+        gpus = [GPUDevice("TEST_DEVICE_NAME", 0, "TEST_PCI_BUS_ID", "TEST_UUID")]
+
+        frequency = 1
+        monitoring_time = 0.1
+        metrics = [CPUAvailableRAM, CPUUsedRAM]
+
+        server = TritonServerFactory.create_server_local(
+            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
+        )
+
+        # Start triton and monitor
+        server.start()
+        cpu_monitor = CPUMonitor(server, frequency, metrics)
+        cpu_monitor.start_recording_metrics()
+        time.sleep(monitoring_time)
+        records = cpu_monitor.stop_recording_metrics()
+
+        # Assert library calls
+        self.server_local_mock.assert_cpu_stats_called()
+
+        # Assert instance types
+        for record in records:
+            self.assertIsInstance(record.value(), float)
+            self.assertIsInstance(record.timestamp(), int)
+
+        # The number of records should be dividable by number of metrics
+        self.assertEqual(len(records) % len(metrics), 0)
+        self.assertGreater(len(records), 0)
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            cpu_monitor.stop_recording_metrics()
+
+        cpu_monitor.destroy()
+        server.stop()
+
+    def test_monitor_disable(self):
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+        gpus = [GPUDevice("TEST_DEVICE_NAME", 0, "TEST_PCI_BUS_ID", "TEST_UUID")]
+
+        frequency = 1
+        monitoring_time = 0.1
+        metrics = []
+
+        server = TritonServerFactory.create_server_local(
+            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
+        )
+
+        # Start triton and monitor
+        server.start()
+        cpu_monitor = CPUMonitor(server, frequency, metrics)
+        cpu_monitor.start_recording_metrics()
+        time.sleep(monitoring_time)
+        cpu_monitor.stop_recording_metrics()
+
+        # Assert no library calls
+        self.server_local_mock.assert_cpu_stats_not_called()
+
+        cpu_monitor.destroy()
+        server.stop()
+
+    def tearDown(self):
+        patch.stopall()
+        self.server_local_mock.stop()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_dcgm_monitor.py b/src/c++/perf_analyzer/tests/test_dcgm_monitor.py
new file mode 100755
index 000000000..ebdbee037
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_dcgm_monitor.py
@@ -0,0 +1,155 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import time
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.device.gpu_device import GPUDevice
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.monitor.dcgm.dcgm_monitor import DCGMMonitor
+from model_analyzer.record.types.gpu_free_memory import GPUFreeMemory
+from model_analyzer.record.types.gpu_power_usage import GPUPowerUsage
+from model_analyzer.record.types.gpu_used_memory import GPUUsedMemory
+from model_analyzer.record.types.gpu_utilization import GPUUtilization
+
+from .common import test_result_collector as trc
+from .mocks.mock_dcgm import MockDCGM
+from .mocks.mock_dcgm_agent import TEST_PCI_BUS_ID, TEST_UUID
+from .mocks.mock_dcgm_field_group_watcher import TEST_RECORD_VALUE
+from .mocks.mock_numba import MockNumba
+
+TEST_DEVICE_NAME = "TEST_DEVICE_NAME"
+TEST_DEVICE_ID = 0
+
+
+class TestDCGMMonitor(trc.TestResultCollector):
+    def setUp(self):
+        self.mock_dcgm = MockDCGM()
+        self.mock_numba = MockNumba(
+            mock_paths=[
+                "model_analyzer.device.gpu_device_factory",
+            ]
+        )
+        self.mock_dcgm.start()
+        self.mock_numba.start()
+
+        self._gpus = [
+            GPUDevice(TEST_DEVICE_NAME, TEST_DEVICE_ID, TEST_PCI_BUS_ID, TEST_UUID)
+        ]
+
+    def test_record_memory(self):
+        # One measurement every 0.01 seconds
+        frequency = 1
+        monitoring_time = 0.1
+        metrics = [GPUUsedMemory, GPUFreeMemory]
+        dcgm_monitor = DCGMMonitor(self._gpus, frequency, metrics)
+        dcgm_monitor.start_recording_metrics()
+        time.sleep(monitoring_time)
+        records = dcgm_monitor.stop_recording_metrics()
+
+        # Assert instance types
+        for record in records:
+            self.assertIsInstance(record.device_uuid(), str)
+            self.assertIsInstance(record.value(), float)
+            self.assertEqual(record.value(), TEST_RECORD_VALUE)
+            self.assertIsInstance(record.timestamp(), int)
+
+        # The number of records should be dividable by number of metrics
+        self.assertEqual(len(records) % len(metrics), 0)
+        self.assertGreater(len(records), 0)
+        self.assertGreaterEqual(
+            records[-1].timestamp() - records[0].timestamp(), monitoring_time
+        )
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            dcgm_monitor.stop_recording_metrics()
+
+        dcgm_monitor.destroy()
+
+        metrics = ["UndefinedTag"]
+        with self.assertRaises(TritonModelAnalyzerException):
+            DCGMMonitor(self._gpus, frequency, metrics)
+
+    def test_record_power(self):
+        # One measurement every 0.01 seconds
+        frequency = 1
+        monitoring_time = 1.1
+        metrics = [GPUPowerUsage]
+        dcgm_monitor = DCGMMonitor(self._gpus, frequency, metrics)
+        dcgm_monitor.start_recording_metrics()
+        time.sleep(monitoring_time)
+        records = dcgm_monitor.stop_recording_metrics()
+
+        # Assert instance types
+        for record in records:
+            self.assertIsInstance(record.device_uuid(), str)
+            self.assertIsInstance(record.value(), float)
+            self.assertEqual(record.value(), TEST_RECORD_VALUE)
+            self.assertIsInstance(record.timestamp(), int)
+
+        # The number of records should be dividable by number of metrics
+        self.assertEqual(len(records) % len(metrics), 0)
+        self.assertGreater(len(records), 0)
+        self.assertGreaterEqual(
+            records[-1].timestamp() - records[0].timestamp(), monitoring_time
+        )
+
+        dcgm_monitor.destroy()
+
+    def test_record_utilization(self):
+        # One measurement every 0.01 seconds
+        frequency = 1
+        monitoring_time = 1.1
+        metrics = [GPUUtilization]
+        dcgm_monitor = DCGMMonitor(self._gpus, frequency, metrics)
+        dcgm_monitor.start_recording_metrics()
+        time.sleep(monitoring_time)
+        records = dcgm_monitor.stop_recording_metrics()
+
+        # Assert instance types
+        for record in records:
+            self.assertIsInstance(record.device_uuid(), str)
+            self.assertIsInstance(record.value(), float)
+            self.assertLessEqual(record.value(), 100)
+            self.assertEqual(record.value(), TEST_RECORD_VALUE)
+            self.assertIsInstance(record.timestamp(), int)
+
+        # The number of records should be dividable by number of metrics
+        self.assertEqual(len(records) % len(metrics), 0)
+        self.assertGreater(len(records), 0)
+        self.assertGreaterEqual(
+            records[-1].timestamp() - records[0].timestamp(), monitoring_time
+        )
+
+        dcgm_monitor.destroy()
+
+    def test_immediate_start_stop(self):
+        frequency = 1
+        metrics = [GPUUsedMemory, GPUFreeMemory]
+        dcgm_monitor = DCGMMonitor(self._gpus, frequency, metrics)
+        dcgm_monitor.start_recording_metrics()
+        dcgm_monitor.stop_recording_metrics()
+        dcgm_monitor.destroy()
+
+    def tearDown(self):
+        patch.stopall()
+        self.mock_dcgm.stop()
+        self.mock_numba.stop()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_ensemble_report_manager.py b/src/c++/perf_analyzer/tests/test_ensemble_report_manager.py
new file mode 100755
index 000000000..a4f04cc89
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_ensemble_report_manager.py
@@ -0,0 +1,178 @@
+#!/usr/bin/env python3
+
+# Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import shutil
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.reports.report_manager import ReportManager
+from model_analyzer.result.result_manager import ResultManager
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+
+from .common import test_result_collector as trc
+from .common.test_utils import ROOT_DIR, evaluate_mock_config
+
+
+class TestEnsembleReportManager(trc.TestResultCollector):
+    def _init_managers(
+        self,
+        models="test_model",
+        num_configs_per_model=10,
+        mode="online",
+        subcommand="profile",
+    ):
+        args = ["model-analyzer", subcommand, "-f", "path-to-config-file"]
+        args.extend(["--checkpoint-directory", f"{ROOT_DIR}/ensemble-ckpt"])
+
+        if subcommand == "profile":
+            args.extend(["--profile-models", models])
+            args.extend(["--model-repository", "/tmp"])
+        else:
+            args.extend(["--report-model-configs", models])
+
+        yaml_str = f"""
+            client_protocol: grpc
+            export_path: {ROOT_DIR}
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand=subcommand)
+        state_manager = AnalyzerStateManager(config=config, server=None)
+
+        state_manager.load_checkpoint(checkpoint_required=True)
+        gpu_info = {"gpu_uuid": {"name": "fake_gpu_name", "total_memory": 1024000000}}
+        self.result_manager = ResultManager(
+            config=config, state_manager=state_manager, constraint_manager=MagicMock()
+        )
+        self.report_manager = ReportManager(
+            mode=mode,
+            config=config,
+            gpu_info=gpu_info,
+            result_manager=self.result_manager,
+            constraint_manager=MagicMock(),
+        )
+
+    def test_ensemble_summary(self):
+        """
+        Ensures the summary report sentence and table are accurate for a basic ensemble model (loaded from a checkpoint)
+        """
+        self._init_managers(models="ensemble_add_sub", subcommand="profile")
+
+        self.report_manager.create_summaries()
+
+        expected_summary_sentence = (
+            "In 21 measurements across 5 configurations, <strong>ensemble_add_sub_config_1</strong> "
+            "is <strong>1%</strong> better than the default configuration at maximizing throughput, "
+            "under the given constraints, on GPU(s) TITAN RTX.<BR><BR><strong>ensemble_add_sub_config_1</strong> is comprised of the following composing models: "
+            "<UL> <LI> <strong>add_config_0</strong>: "
+            "1 GPU instance with a max batch size of 0 on platform python </LI><LI> <strong>sub_config_1</strong>: "
+            "2 GPU instances with a max batch size of 0 on platform python </LI> </UL>"
+        )
+
+        summary_table, summary_sentence = self.report_manager._build_summary_table(
+            report_key="ensemble_add_sub",
+            num_measurements=21,
+            num_configurations=5,
+            gpu_name="TITAN RTX",
+            report_gpu_metrics=True,
+        )
+
+        self.assertEqual(summary_sentence, expected_summary_sentence)
+
+        # Check the first row of the table
+        self.assertEqual(
+            summary_table._rows[0][0], "ensemble_add_sub_config_1"
+        )  # model config name
+        self.assertEqual(summary_table._rows[0][1], "(0, 0)")  # max batch size
+        self.assertEqual(
+            summary_table._rows[0][2], "(Disabled, Disabled)"
+        )  # dynamic batching
+        self.assertEqual(summary_table._rows[0][3], "(1:GPU, 2:GPU)")  # instance count
+        self.assertEqual(summary_table._rows[0][4], "3.655")  # p99 latency
+        self.assertEqual(summary_table._rows[0][5], "2865.88")  # throughput
+        self.assertEqual(summary_table._rows[0][6], 870)  # max gpu memory
+        self.assertEqual(summary_table._rows[0][7], 0.0)  # GPU utilization
+
+    def test_ensemble_summary_cpu_only(self):
+        """
+        Ensures the summary report sentence and table are accurate for a basic ensemble model (loaded from a checkpoint)
+        when the cpu only flag is set
+        """
+
+        self._init_managers(models="ensemble_add_sub", subcommand="profile")
+
+        self.report_manager.create_summaries()
+
+        expected_summary_sentence = (
+            "In 21 measurements across 5 configurations, <strong>ensemble_add_sub_config_1</strong> "
+            "is <strong>1%</strong> better than the default configuration at maximizing throughput, "
+            "under the given constraints.<BR><BR><strong>ensemble_add_sub_config_1</strong> is comprised of the following composing models: "
+            "<UL> <LI> <strong>add_config_0</strong>: "
+            "1 GPU instance with a max batch size of 0 on platform python </LI><LI> <strong>sub_config_1</strong>: "
+            "2 GPU instances with a max batch size of 0 on platform python </LI> </UL>"
+        )
+
+        summary_table, summary_sentence = self.report_manager._build_summary_table(
+            report_key="ensemble_add_sub",
+            num_measurements=21,
+            num_configurations=5,
+            gpu_name="TITAN RTX",
+            report_gpu_metrics=False,
+        )
+
+        self.assertEqual(summary_sentence, expected_summary_sentence)
+
+        # Check the first row of the table
+        self.assertEqual(
+            summary_table._rows[0][0], "ensemble_add_sub_config_1"
+        )  # model config name
+        self.assertEqual(summary_table._rows[0][1], "0, 0")  # max batch size
+        self.assertEqual(
+            summary_table._rows[0][2], "(Disabled, Disabled)"
+        )  # dynamic batching
+        self.assertEqual(summary_table._rows[0][3], "1:GPU, 2:GPU")  # instance count
+        self.assertEqual(summary_table._rows[0][4], "3.655")  # p99 latency
+        self.assertEqual(summary_table._rows[0][5], "2865.88")  # throughput
+
+    def test_ensemble_detailed(self):
+        """
+        Ensures the detailed report sentence is accurate for a basic ensemble model (loaded from a checkpoint)
+        """
+        self._init_managers(models="ensemble_add_sub_config_1", subcommand="report")
+
+        self.report_manager._add_detailed_report_data()
+        self.report_manager._build_detailed_table("ensemble_add_sub_config_1")
+        detailed_sentence = self.report_manager._build_detailed_info(
+            "ensemble_add_sub_config_1"
+        )
+
+        expected_detailed_sentence = (
+            "<strong>ensemble_add_sub_config_1</strong> is comprised of the "
+            "following composing models:<LI> <strong>add_config_0</strong>: "
+            "1 GPU instance with a max batch size of 0 on platform python </LI><LI> "
+            "<strong>sub_config_1</strong>: 2 GPU instances with a max batch size of 0 "
+            "on platform python </LI><br>5 measurement(s) were obtained for the model config "
+            "on GPU(s)  with total memory 0 GB."
+        )
+
+        self.assertEqual(detailed_sentence, expected_detailed_sentence)
+
+    def tearDown(self):
+        patch.stopall()
+        shutil.rmtree(f"{ROOT_DIR}/reports")
+        shutil.rmtree(f"{ROOT_DIR}/plots")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_file_writer.py b/src/c++/perf_analyzer/tests/test_file_writer.py
new file mode 100755
index 000000000..f86ec393a
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_file_writer.py
@@ -0,0 +1,67 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.output.file_writer import FileWriter
+
+from .common import test_result_collector as trc
+from .mocks.mock_io import MockIOMethods
+
+TEST_FILENAME = "test_filename"
+
+
+class TestFileWriterMethods(trc.TestResultCollector):
+    def setUp(self):
+        self._mock_path = "model_analyzer.output.file_writer"
+        self.io_mock = MockIOMethods(mock_paths=["model_analyzer.output.file_writer"])
+        self.io_mock.start()
+
+    def test_write(self):
+        # Create and use writer
+        writer = FileWriter(filename=TEST_FILENAME)
+
+        # Check for exception on open and write
+        self.io_mock.raise_exception_on_open()
+        err_str = "Expected TritonModelAnalyzerException on malformed input."
+        with self.assertRaises(TritonModelAnalyzerException, msg=err_str):
+            writer.write("test")
+        self.io_mock.reset()
+
+        self.io_mock.raise_exception_on_write()
+        err_str = "Expected TritonModelAnalyzerException on malformed input."
+        with self.assertRaises(TritonModelAnalyzerException, msg=err_str):
+            writer.write("test")
+        self.io_mock.reset()
+
+        # Check mock call on successful write
+        writer.write("test")
+        self.io_mock.assert_write_called_with_args(self._mock_path, "test")
+
+        # Perform checks for stdout
+        writer = FileWriter()
+        writer.write("test")
+        self.io_mock.assert_print_called_with_args(self._mock_path, "test")
+
+    def tearDown(self):
+        patch.stopall()
+        self.io_mock.stop()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_html_report.py b/src/c++/perf_analyzer/tests/test_html_report.py
new file mode 100755
index 000000000..9473ede9c
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_html_report.py
@@ -0,0 +1,167 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import base64
+import unittest
+from unittest.mock import mock_open, patch
+
+from model_analyzer.reports.html_report import HTMLReport
+from model_analyzer.result.result_table import ResultTable
+
+from .common import test_result_collector as trc
+
+
+class TestHTMLReportMethods(trc.TestResultCollector):
+    """
+    Tests the methods of the HTMLReport class
+    """
+
+    def setUp(self):
+        self.maxDiff = None
+        self.report = HTMLReport()
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_add_title(self):
+        self.report.add_title("Test HTML Report")
+        expected_report_body = "<html><head><style></style></head><body><center><h1>Test HTML Report</h1></center></body></html>"
+        self.assertEqual(self.report.document(), expected_report_body)
+
+    def test_add_subheading(self):
+        # Add one subheading
+        self.report.add_subheading("Throughput vs. Latency")
+        expected_report_body = "<html><head><style></style></head><body><h3>Throughput vs. Latency</h3></body></html>"
+        self.assertEqual(self.report.document(), expected_report_body)
+
+        # Add another subheading
+        self.report.add_subheading("GPU Memory vs. Latency")
+        expected_report_body = (
+            "<html><head><style></style></head><body><h3>Throughput vs. Latency</h3>"
+            "<h3>GPU Memory vs. Latency</h3></body></html>"
+        )
+        self.assertEqual(self.report.document(), expected_report_body)
+
+    def test_add_image(self):
+        with patch(
+            "model_analyzer.reports.html_report.open",
+            mock_open(read_data=bytes(">:(".encode("ascii"))),
+        ):
+            self.report.add_images(["test_image_file"], ["test_caption"])
+        img_content = base64.b64encode(bytes(">:(".encode("ascii"))).decode("ascii")
+        expected_report_body = (
+            f'<html><head><style></style></head><body><center><div><div class="image" style="float:center;width:100%"><img src="data:image/png;base64,{img_content}"'
+            ' style="width:100%"><center><div style="font-weight:bold;font-size:12;padding-bottom:20px">'
+            "test_caption</div></center></div></div></center></body></html>"
+        )
+        self.assertEqual(self.report.document(), expected_report_body)
+
+    def test_add_paragraph(self):
+        test_paragraph = (
+            "This is a test paragraph with a lot to say."
+            " There is more than one line in this paragraph."
+        )
+
+        # Default font size
+        self.report.add_paragraph(test_paragraph)
+        expected_report_body = (
+            '<html><head><style></style></head><body><div style="font-size:14">'
+            f"<p>{test_paragraph}</p></div></body></html>"
+        )
+        self.assertEqual(self.report.document(), expected_report_body)
+
+        # custom font size
+        self.report.add_paragraph(test_paragraph, font_size=20)
+        expected_report_body = (
+            '<html><head><style></style></head><body><div style="font-size:14">'
+            f'<p>{test_paragraph}</p></div><div style="font-size:20">'
+            f"<p>{test_paragraph}</p></div></body></html>"
+        )
+        self.assertEqual(self.report.document(), expected_report_body)
+
+    def test_add_table(self):
+        result_table = ResultTable(["header1", "header2"])
+        # Try empty table
+        self.report.add_table(table=result_table)
+        table_style = "border: 1px solid black;border-collapse: collapse;text-align: center;width: 80%;padding: 5px 10px;font-size: 11pt"
+        expected_report_body = (
+            "<html><head><style></style></head><body>"
+            f'<center><table style="{table_style}">'
+            "<tr>"
+            f'<th style="{table_style}">header1</th>'
+            f'<th style="{table_style}">header2</th>'
+            "</tr>"
+            "</table></center>"
+            "</body></html>"
+        )
+        self.assertEqual(self.report.document(), expected_report_body)
+
+        # Fill table
+        for i in range(2):
+            result_table.insert_row_by_index([f"value{i}1", f"value{i}2"])
+
+        # Table has 5 rows
+        self.report.add_table(table=result_table)
+        expected_report_body = (
+            "<html><head><style></style></head><body>"
+            f'<center><table style="{table_style}">'
+            "<tr>"
+            f'<th style="{table_style}">header1</th>'
+            f'<th style="{table_style}">header2</th>'
+            "</tr>"
+            "</table></center>"
+            f'<center><table style="{table_style}">'
+            "<tr>"
+            f'<th style="{table_style}">header1</th>'
+            f'<th style="{table_style}">header2</th>'
+            "</tr>"
+            "<tr>"
+            f'<td style="{table_style}">value01</td>'
+            f'<td style="{table_style}">value02</td>'
+            "</tr>"
+            "<tr>"
+            f'<td style="{table_style}">value11</td>'
+            f'<td style="{table_style}">value12</td>'
+            "</tr>"
+            "</table></center>"
+            "</body></html>"
+        )
+
+        self.assertEqual(self.report.document(), expected_report_body)
+
+    def test_write_report(self):
+        self.report.add_title("Test HTML Report")
+        self.report.add_subheading("Throughput vs. Latency")
+        test_paragraph = (
+            "This is a test paragraph with a lot to say."
+            " There is more than one line in this paragraph."
+        )
+        self.report.add_paragraph(test_paragraph, font_size=14)
+
+        expected_report_body = (
+            "<html><head><style></style></head><body><center><h1>Test HTML Report</h1></center><h3>Throughput vs. Latency</h3>"
+            f'<div style="font-size:14"><p>{test_paragraph}</p></div>'
+            "</body></html>"
+        )
+
+        with patch("builtins.open", mock_open()) as mocked_file:
+            self.report.write_report("test_report_filename")
+        mocked_file.assert_called_once_with(f"test_report_filename", "w")
+        mocked_file().write.assert_called_once_with(f"{expected_report_body}")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_logger.py b/src/c++/perf_analyzer/tests/test_logger.py
new file mode 100755
index 000000000..fec7c1e88
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_logger.py
@@ -0,0 +1,175 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# This test exercises all of the cli options.
+# It does a basic test of the cli and help message.
+# Then the rest of the testing uses an OptionStruct, which holds all
+# of the data necessary to test a command, and feeds that to the
+# CLI parser. The result of the CLI parsing is compared against the
+# expected value for the CLI. Default values are also verified as well as
+# values that are expected to cause failures.
+
+import logging
+import re
+import sys
+import unittest
+from contextlib import contextmanager
+from io import StringIO
+from unittest.mock import patch
+
+from model_analyzer.constants import LOGGER_NAME
+from model_analyzer.log_formatter import setup_logging
+
+from .common import test_result_collector as trc
+
+logger = logging.getLogger(LOGGER_NAME)
+
+
+@contextmanager
+def captured_output():
+    new_out, new_err = StringIO(), StringIO()
+    old_out, old_err = sys.stdout, sys.stderr
+    try:
+        sys.stdout, sys.stderr = new_out, new_err
+        yield sys.stdout, sys.stderr
+    finally:
+        sys.stdout, sys.stderr = old_out, old_err
+
+
+class TestLogger(trc.TestResultCollector):
+    def tearDown(self):
+        patch.stopall()
+
+    def test_info_normal(self):
+        """Test expected format of logger.info"""
+        with captured_output() as (out, err):
+            setup_logging(verbose=False, quiet=False)
+            logger.info("ABC")
+            expected_output = "[Model Analyzer] ABC"
+            output = out.getvalue().strip()
+            self.assertEqual(output, expected_output)
+
+    def test_info_quiet(self):
+        """
+        Test expected format of logger.info in quiet mode
+        (Nothing should be printed)
+        """
+        with captured_output() as (out, err):
+            setup_logging(verbose=False, quiet=True)
+            logger.info("ABC")
+            expected_output = ""
+            output = out.getvalue().strip()
+            self.assertEqual(output, expected_output)
+
+    def test_info_verbose(self):
+        """
+        Test expected format of logger.info in verbose mode
+        (Extra date info should be added to the start of the line)
+        """
+        with captured_output() as (out, err):
+            setup_logging(verbose=True, quiet=False)
+            logger.info("ABC")
+            output = out.getvalue().strip()
+            self.assertTrue(re.match("^\d\d:\d\d:\d\d \[Model Analyzer\] ABC$", output))
+
+    def test_debug_normal(self):
+        """
+        Test expected format of logger.debug
+        (Nothing should be printed since we aren't in verbose mode)
+        """
+        with captured_output() as (out, err):
+            setup_logging(verbose=False, quiet=False)
+            logger.debug("ABC")
+            expected_output = ""
+            output = out.getvalue().strip()
+            self.assertEqual(output, expected_output)
+
+    def test_debug_quiet(self):
+        """
+        Test expected format of logger.debug in quiet mode
+        (Nothing should be printed since we aren't in verbose mode)
+        """
+        with captured_output() as (out, err):
+            setup_logging(verbose=False, quiet=True)
+            logger.debug("ABC")
+            expected_output = ""
+            output = out.getvalue().strip()
+            self.assertEqual(output, expected_output)
+
+    def test_debug_verbose(self):
+        """
+        Test expected format of logger.debug in verbose mode
+        (Extra date info should be added to the start of the line)
+        """
+        with captured_output() as (out, err):
+            setup_logging(verbose=True, quiet=False)
+            logger.debug("ABC")
+            output = out.getvalue().strip()
+            self.assertTrue(
+                re.match("^\d\d:\d\d:\d\d \[Model Analyzer\] DEBUG: ABC$", output)
+            )
+
+    def test_error_normal(self):
+        """
+        Test expected format of logger.error
+        """
+        with captured_output() as (out, err):
+            setup_logging(verbose=False, quiet=False)
+            logger.error("ABC")
+            expected_output = "[Model Analyzer] ERROR: ABC"
+            output = out.getvalue().strip()
+            self.assertEqual(output, expected_output)
+
+    def test_error_quiet(self):
+        """
+        Test expected format of logger.error in quiet mode
+        (It will be printed despite verbose mode)
+        """
+        with captured_output() as (out, err):
+            setup_logging(verbose=False, quiet=True)
+            logger.error("ABC")
+            expected_output = "[Model Analyzer] ERROR: ABC"
+            output = out.getvalue().strip()
+            self.assertEqual(output, expected_output)
+
+    def test_error_verbose(self):
+        """
+        Test expected format of logger.error in verbose mode
+        (Extra date info should be added to the start of the line)
+        """
+        with captured_output() as (out, err):
+            setup_logging(verbose=True, quiet=False)
+            logger.error("ABC")
+            output = out.getvalue().strip()
+            self.assertTrue(
+                re.match("^\d\d:\d\d:\d\d \[Model Analyzer\] ERROR: ABC$", output)
+            )
+
+    def test_quiet_trumps_verbose(self):
+        """
+        Test that calling setup_logging with both verbose and quiet true
+        will result in quiet mode
+        """
+        with captured_output() as (out, err):
+            setup_logging(verbose=True, quiet=True)
+            logger.info("ABC")
+            expected_output = ""
+            output = out.getvalue().strip()
+            self.assertEqual(output, expected_output)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_config.py b/src/c++/perf_analyzer/tests/test_model_config.py
new file mode 100755
index 000000000..5371107dc
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_model_config.py
@@ -0,0 +1,472 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import unittest
+from unittest.mock import MagicMock, mock_open, patch
+
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.triton.model.model_config import ModelConfig
+
+from .common import test_result_collector as trc
+from .mocks.mock_model_config import MockModelConfig
+
+
+class TestModelConfig(trc.TestResultCollector):
+    def setUp(self):
+        self._model_config = {
+            "name": "classification_chestxray_v1",
+            "platform": "tensorflow_graphdef",
+            "max_batch_size": 32,
+            "input": [
+                {
+                    "name": "NV_MODEL_INPUT",
+                    "data_type": "TYPE_FP32",
+                    "format": "FORMAT_NHWC",
+                    "dims": ["256", "256", "3"],
+                }
+            ],
+            "output": [
+                {
+                    "name": "NV_MODEL_OUTPUT",
+                    "data_type": "TYPE_FP32",
+                    "dims": ["15"],
+                    "label_filename": "chestxray_labels.txt",
+                }
+            ],
+            "instance_group": [{"count": 1, "kind": "KIND_GPU"}],
+        }
+
+        # Equivalent protobuf for the model config above.
+        self._model_config_protobuf = """
+name: "classification_chestxray_v1"
+platform: "tensorflow_graphdef"
+max_batch_size: 32
+input [
+  {
+    name: "NV_MODEL_INPUT"
+    data_type: TYPE_FP32
+    format: FORMAT_NHWC
+    dims: [256, 256, 3]
+  }
+]
+output [
+  {
+    name: "NV_MODEL_OUTPUT"
+    data_type: TYPE_FP32
+    dims: [15]
+    label_filename: "chestxray_labels.txt"
+  }
+]
+instance_group [
+  {
+    kind: KIND_GPU
+    count: 1
+  }
+]
+"""
+
+        # Ensemble model config
+        self._ensemble_model_config_protobuf = """
+    name: "ensemble_python_resnet50"
+platform: "ensemble"
+max_batch_size: 256
+input [
+  {
+    name: "INPUT"
+    data_type: TYPE_UINT8
+    dims: [ -1 ]
+  }
+]
+output [
+  {
+    name: "OUTPUT"
+    data_type: TYPE_FP32
+    dims: [ 1000 ]
+  }
+]
+ensemble_scheduling {
+  step [
+    {
+      model_name: "preprocess"
+      model_version: -1
+      input_map {
+        key: "INPUT_0"
+        value: "INPUT"
+      }
+      output_map {
+        key: "OUTPUT_0"
+        value: "preprocessed_image"
+      }
+    },
+    {
+      model_name: "resnet50_trt"
+      model_version: -1
+      input_map {
+        key: "input"
+        value: "preprocessed_image"
+      }
+      output_map {
+        key: "output"
+        value: "OUTPUT"
+      }
+    }
+  ]
+}
+"""
+
+        self._ensemble_model_config_protobuf_no_scheduling = """
+    name: "ensemble_python_resnet50"
+platform: "ensemble"
+max_batch_size: 256
+input [
+  {
+    name: "INPUT"
+    data_type: TYPE_UINT8
+    dims: [ -1 ]
+  }
+]
+output [
+  {
+    name: "OUTPUT"
+    data_type: TYPE_FP32
+    dims: [ 1000 ]
+  }
+]
+"""
+
+        self._ensemble_model_config_protobuf_no_step = """
+    name: "ensemble_python_resnet50"
+platform: "ensemble"
+max_batch_size: 256
+input [
+  {
+    name: "INPUT"
+    data_type: TYPE_UINT8
+    dims: [ -1 ]
+  }
+]
+output [
+  {
+    name: "OUTPUT"
+    data_type: TYPE_FP32
+    dims: [ 1000 ]
+  }
+]
+ensemble_scheduling {
+}
+"""
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_create_from_file(self):
+        test_protobuf = self._model_config_protobuf
+        mock_model_config = MockModelConfig(test_protobuf)
+        mock_model_config.start()
+        model_config = ModelConfig._create_from_file("/path/to/model_config")
+        self.assertEqual(model_config.get_config(), self._model_config)
+        mock_model_config.stop()
+
+    def test_create_from_dict(self):
+        model_config = ModelConfig.create_from_dictionary(self._model_config)
+        self.assertEqual(model_config.get_config(), self._model_config)
+
+        new_config = {"instance_group": [{"count": 2, "kind": "KIND_CPU"}]}
+        model_config.set_config(new_config)
+        self.assertEqual(model_config.get_config(), new_config)
+
+    def test_write_config_file(self):
+        model_config = ModelConfig.create_from_dictionary(self._model_config)
+        model_output_path = os.path.abspath("./model_config")
+
+        mock_model_config = MockModelConfig()
+        mock_model_config.start()
+        # Write the model config to output
+        with patch(
+            "model_analyzer.triton.model.model_config.open", mock_open()
+        ) as mocked_file:
+            with patch(
+                "model_analyzer.triton.model.model_config.copytree", MagicMock()
+            ):
+                model_config.write_config_to_file(model_output_path, "/mock/path", None)
+            content = mocked_file().write.call_args.args[0]
+        mock_model_config.stop()
+
+        mock_model_config = MockModelConfig(content)
+        mock_model_config.start()
+        model_config_from_file = ModelConfig._create_from_file(model_output_path)
+        self.assertEqual(model_config_from_file.get_config(), self._model_config)
+        mock_model_config.stop()
+
+        # output path doesn't exist
+        with patch(
+            "model_analyzer.triton.model.model_config.os.path.exists",
+            MagicMock(return_value=False),
+        ):
+            with self.assertRaises(TritonModelAnalyzerException):
+                ModelConfig._create_from_file(model_output_path)
+
+        # output path is a file
+        with patch(
+            "model_analyzer.triton.model.model_config.os.path.isfile",
+            MagicMock(return_value=True),
+        ):
+            with self.assertRaises(TritonModelAnalyzerException):
+                ModelConfig._create_from_file(model_output_path)
+
+    @patch(
+        "model_analyzer.triton.model.model_config.os.listdir",
+        MagicMock(return_value=["1", "config.pbtxt", "output0_labels.txt"]),
+    )
+    @patch("model_analyzer.triton.model.model_config.copytree")
+    @patch("model_analyzer.triton.model.model_config.os.symlink")
+    def test_write_config_to_file_with_relative_path(self, mock_os_symlink, *args):
+        """
+        Tests that the call to os.symlink() within write_config_to_file() uses
+        a valid relative path when user uses a relative path with the
+        `--output-model-repository-path` option
+        """
+
+        model_config = ModelConfig.create_from_dictionary(self._model_config)
+
+        model_path = "./output_model_repository/model_config_1"
+        src_model_path = "/tmp/src_model_repository/model"
+        last_model_path = "./output_model_repository/model_config_0"
+
+        mock_model_config = MockModelConfig()
+        mock_model_config.start()
+        model_config.write_config_to_file(model_path, src_model_path, last_model_path)
+        mock_model_config.stop()
+
+        mock_os_symlink.assert_any_call(
+            "../model_config_0/1", "./output_model_repository/model_config_1/1"
+        )
+        mock_os_symlink.assert_any_call(
+            "../model_config_0/output0_labels.txt",
+            "./output_model_repository/model_config_1/output0_labels.txt",
+        )
+
+    def test_instance_group_string(self):
+        """Test out all corner cases of instance_group_string()"""
+
+        def _test_helper(config_dict, expected_result, gpu_count=None):
+            model_config = ModelConfig.create_from_dictionary(config_dict)
+            if gpu_count:
+                instance_group_str = model_config.instance_group_string(gpu_count)
+            else:
+                instance_group_str = model_config.instance_group_string()
+            self.assertEqual(instance_group_str, expected_result)
+
+        # No instance group info in model_config_dict:
+        #  - default to 1 per GPU
+        model_config_dict = {}
+        _test_helper(model_config_dict, "1:GPU", gpu_count=1)
+
+        # No instance group info in model_config_dict:
+        #  - 1 per GPU -- if 2 gpus then 2 total
+        model_config_dict = {}
+        _test_helper(model_config_dict, "2:GPU", gpu_count=2)
+
+        # No instance group info in model_config_dict:
+        #  - default to 1 on CPU if cuda not available
+        model_config_dict = {}
+        with patch("numba.cuda.is_available", MagicMock(return_value=False)):
+            _test_helper(model_config_dict, "1:CPU", gpu_count=5)
+
+        # 2 per GPU, 3 gpus in the system = 6 total
+        model_config_dict = {
+            "instance_group": [
+                {
+                    "count": 2,
+                    "kind": "KIND_GPU",
+                }
+            ]
+        }
+        _test_helper(model_config_dict, "6:GPU", gpu_count=3)
+
+        # 1 on GPU0 only = 1 total despite 2 GPUs in the system
+        model_config_dict = {
+            "instance_group": [{"count": 1, "kind": "KIND_GPU", "gpus": [0]}]
+        }
+        _test_helper(model_config_dict, "1:GPU", gpu_count=2)
+
+        # 1 on ALL gpus + 2 each on [1 and 3] + 3 more on CPUs
+        # with 4 GPUs in the system:
+        #   8 on GPU and 3 on CPU
+        model_config_dict = {
+            "instance_group": [
+                {"count": 1, "kind": "KIND_GPU"},
+                {"count": 2, "kind": "KIND_GPU", "gpus": [1, 3]},
+                {"count": 3, "kind": "KIND_CPU"},
+            ]
+        }
+        _test_helper(model_config_dict, "8:GPU + 3:CPU", gpu_count=4)
+
+    def test_instance_group_count(self):
+        """Test out all corner cases of instance_group_count()"""
+
+        def _test_helper(config_dict, expected_result, gpu_count=None):
+            model_config = ModelConfig.create_from_dictionary(config_dict)
+            instance_group_count = model_config.instance_group_count(gpu_count)
+            self.assertEqual(instance_group_count, expected_result)
+
+        # No instance group info in model_config_dict:
+        #  - default to 1 per GPU
+        model_config_dict = {}
+        _test_helper(model_config_dict, 1, gpu_count=1)
+
+        # No instance group info in model_config_dict:
+        #  - 1 per GPU -- if 2 gpus then 2 total
+        model_config_dict = {}
+        _test_helper(model_config_dict, 2, gpu_count=2)
+
+        # 2 per GPU, 3 gpus in the system = 6 total
+        model_config_dict = {
+            "instance_group": [
+                {
+                    "count": 2,
+                    "kind": "KIND_GPU",
+                }
+            ]
+        }
+        _test_helper(model_config_dict, 6, gpu_count=3)
+
+        # 1 on ALL gpus + 2 each on [1 and 3] + 3 more on CPUs
+        # with 4 GPUs in the system:
+        #   8 on GPU and 3 on CPU
+        model_config_dict = {
+            "instance_group": [
+                {"count": 1, "kind": "KIND_GPU"},
+                {"count": 2, "kind": "KIND_GPU", "gpus": [1, 3]},
+                {"count": 3, "kind": "KIND_CPU"},
+            ]
+        }
+        _test_helper(model_config_dict, 11, gpu_count=4)
+
+    def test_is_ensemble(self):
+        """Test that we recognize if the platform is ensemble"""
+
+        tensor_protobuf = self._model_config_protobuf
+        mock_model_config = MockModelConfig(tensor_protobuf)
+        mock_model_config.start()
+        model_config = ModelConfig._create_from_file("/path/to/model_config")
+        self.assertFalse(model_config.is_ensemble())
+        mock_model_config.stop()
+
+        ensemble_protobuf = self._ensemble_model_config_protobuf
+        mock_model_config = MockModelConfig(ensemble_protobuf)
+        mock_model_config.start()
+        model_config = ModelConfig._create_from_file("/path/to/model_config")
+        self.assertTrue(model_config.is_ensemble())
+        mock_model_config.stop()
+
+    def test_ensemble_composing_models(self):
+        """Test that we can extract an ensembles composing model names"""
+
+        ensemble_protobuf = self._ensemble_model_config_protobuf
+        mock_model_config = MockModelConfig(ensemble_protobuf)
+        mock_model_config.start()
+        model_config = ModelConfig._create_from_file("/path/to/model_config")
+
+        self.assertEqual(
+            model_config.get_ensemble_composing_models(), ["preprocess", "resnet50_trt"]
+        )
+        mock_model_config.stop()
+
+    def test_set_composing_model_name(self):
+        """Test setting a variant name for an ensemble's composing model"""
+
+        ensemble_protobuf = self._ensemble_model_config_protobuf
+        mock_model_config = MockModelConfig(ensemble_protobuf)
+        mock_model_config.start()
+        model_config = ModelConfig._create_from_file("/path/to/model_config")
+
+        model_config.set_composing_model_variant_name(
+            composing_model_name="preprocess", variant_name="preprocess_config_0"
+        )
+        model_config.set_composing_model_variant_name(
+            composing_model_name="resnet50_trt", variant_name="resnet50_trt_config_1"
+        )
+
+        self.assertEqual(
+            model_config.get_ensemble_composing_models(),
+            ["preprocess_config_0", "resnet50_trt_config_1"],
+        )
+        mock_model_config.stop()
+
+    def test_badly_formed_ensemble_configs(self):
+        """Test cases where protobuf is incorrect for an ensemble platform"""
+
+        # Tests where platform is not ensemble
+        ensemble_protobuf = self._model_config_protobuf
+        mock_model_config = MockModelConfig(ensemble_protobuf)
+        mock_model_config.start()
+        model_config = ModelConfig._create_from_file("/path/to/model_config")
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            model_config.get_ensemble_composing_models()
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            model_config.set_composing_model_variant_name(
+                composing_model_name="preprocess", variant_name="preprocess_config_0"
+            )
+
+        mock_model_config.stop()
+
+        # Tests where platform is ensemble, but scheduling section doesn't exist
+        ensemble_protobuf = self._ensemble_model_config_protobuf_no_scheduling
+        mock_model_config = MockModelConfig(ensemble_protobuf)
+        mock_model_config.start()
+        model_config = ModelConfig._create_from_file("/path/to/model_config")
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            model_config.get_ensemble_composing_models()
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            model_config.set_composing_model_variant_name(
+                composing_model_name="preprocess", variant_name="preprocess_config_0"
+            )
+
+        mock_model_config.stop()
+
+        # Tests where platform is ensemble, scheduling exists, but step is omitted
+        ensemble_protobuf = self._ensemble_model_config_protobuf_no_step
+        mock_model_config = MockModelConfig(ensemble_protobuf)
+        mock_model_config.start()
+        model_config = ModelConfig._create_from_file("/path/to/model_config")
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            model_config.get_ensemble_composing_models()
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            model_config.set_composing_model_variant_name(
+                composing_model_name="preprocess", variant_name="preprocess_config_0"
+            )
+
+        mock_model_config.stop()
+
+    def test_set_model_name(self):
+        """
+        Test that we can set the model name of a config
+        """
+        model_config = ModelConfig.create_from_dictionary(self._model_config)
+        model_config.set_model_name("new_model_name")
+        self.assertEqual(model_config.get_field("name"), "new_model_name")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_config_generator.py b/src/c++/perf_analyzer/tests/test_model_config_generator.py
new file mode 100755
index 000000000..3717c04be
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_model_config_generator.py
@@ -0,0 +1,927 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.generate.base_model_config_generator import (
+    BaseModelConfigGenerator,
+)
+from model_analyzer.config.generate.model_config_generator_factory import (
+    ModelConfigGeneratorFactory,
+)
+from model_analyzer.config.generate.model_profile_spec import ModelProfileSpec
+from model_analyzer.config.generate.model_variant_name_manager import (
+    ModelVariantNameManager,
+)
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.triton.model.model_config import ModelConfig
+
+from .common import test_result_collector as trc
+from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
+from .mocks.mock_model_config import MockModelConfig
+from .mocks.mock_os import MockOSMethods
+
+
+class TestModelConfigGenerator(trc.TestResultCollector):
+    def __init__(self, methodname):
+        super().__init__(methodname)
+        self._fake_throughput = 1
+
+    def test_direct_no_params(self):
+        """
+        Test direct modes with no model_config_parameters specified
+
+        It will just sweep instance count and max_batch_size (with dynamic batching on),
+        and default config (empty dict) will be included
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_direct_no_params_search_disable(self):
+        """
+        Test direct mode with no model_config_parameters specified and run_search disabled
+
+        This will just return a single empty config, since there are no parameters to combine
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_disable: True
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [{'max_batch_size': 8}]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_direct_yes_params_search_disable(self):
+        """
+        Test direct modes with model_config_parameters specified and run_search disabled
+
+        This will just combine all model_config_parameters
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_instance_count: 16
+            run_config_search_disable: True
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        max_batch_size: [1,4,16]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1,2]
+            """)
+
+        expected_configs = [
+            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
+            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
+            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 16},
+            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
+            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
+            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 16}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_run_config_search_options(self):
+        """
+        Test that ModelConfigGenerator will honor run_config_search_max_instance_count
+        and run_config_search_max_model_batch_size and run_config_search_min_model_batch_size
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_instance_count: 3
+            run_config_search_min_model_batch_size: 2
+            run_config_search_max_model_batch_size: 16
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_run_config_search_min_instance_counts(self):
+        """
+        Test that ModelConfigGenerator will honor run_config_search_min_instance_count
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_min_instance_count: 2
+            run_config_search_max_instance_count: 3
+            run_config_search_min_model_batch_size: 2
+            run_config_search_max_model_batch_size: 8
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_non_power_of_two_max_batch_size(self):
+        """
+        Test that ModelConfigGenerator will correctly sweep max_batch_size with
+        input values that aren't a power of 2
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_instance_count: 2
+            run_config_search_min_model_batch_size: 3
+            run_config_search_max_model_batch_size: 15
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'dynamic_batching': {}, 'max_batch_size': 3, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 3, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 6, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 6, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 12, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 12, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_direct_yes_params_specified(self):
+        """
+        Test direct mode with model_config_parameters specified
+
+        It will combine all legal combinations of config values, and
+        default config (None) will be included
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_instance_count: 16
+            run_config_search_disable: False
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        max_batch_size: [1,4,16]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1,2]
+            """)
+
+        expected_configs = [
+            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
+            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
+            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 16},
+            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
+            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
+            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 16}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_direct_cpu_only(self):
+        """
+        Test direct mode with cpu_only=true
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_instance_count: 2
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 16
+            profile_models:
+                - my-model:
+                    cpu_only: True
+            """)
+
+        expected_configs = [
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_CPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_CPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 1, 'kind': 'KIND_CPU'}]},
+            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 2, 'kind': 'KIND_CPU'}]}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_direct_max_batch_size_0(self):
+        """
+        Test direct mode with the the default config stating max_batch_size=0
+
+        max_batch_size and dynamic_batching should not be part of the resulting configs
+        """
+
+        # yapf: disable
+        protobuf = """
+            platform: "fake_platform"
+            max_batch_size: 0
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 1
+            }
+            ]
+            """
+
+        yaml_str = ("""
+            run_config_search_max_instance_count: 4
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'platform': "fake_platform", 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
+
+    def test_direct_max_batch_size_unspecified(self):
+        """
+        Test direct mode with the the default config not specifying max_batch_size
+
+        max_batch_size and dynamic_batching should not be part of the resulting configs
+        """
+
+        # yapf: disable
+        protobuf = """
+            platform: "fake_platform"
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 1
+            }
+            ]
+            """
+
+        yaml_str = ("""
+            run_config_search_max_instance_count: 4
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'platform': "fake_platform", 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
+
+    def test_direct_sequence_batching(self):
+        """
+        Test direct mode with the the default config specifying sequence_batching
+
+        dynamic_batching should not be part of the resulting configs
+        """
+
+        # yapf: disable
+        protobuf = """
+            platform: "fake_platform"
+            max_batch_size: 4
+            sequence_batching {}
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 1
+            }
+            ]
+            """
+
+        yaml_str = ("""
+            run_config_search_max_instance_count: 4
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 16
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'platform': "fake_platform", 'max_batch_size':  8, 'sequence_batching': {}, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'max_batch_size':  8, 'sequence_batching': {}, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'max_batch_size':  8, 'sequence_batching': {}, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'max_batch_size':  8, 'sequence_batching': {}, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'max_batch_size': 16, 'sequence_batching': {}, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'max_batch_size': 16, 'sequence_batching': {}, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'max_batch_size': 16, 'sequence_batching': {}, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'platform': "fake_platform", 'max_batch_size': 16, 'sequence_batching': {}, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
+
+    def test_direct_nonempty_default_config(self):
+        """
+        Test direct mode with the the default config containing some values
+
+        It will keep values that aren't part of the search, and will overwrite
+        any values that are part of the search
+        """
+
+        # yapf: disable
+        protobuf = """
+            platform: "fake_platform"
+            max_batch_size: 4
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 1
+            }
+            ]
+            """
+
+        yaml_str = ("""
+            run_config_search_max_instance_count: 4
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'platform': "fake_platform", 'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'platform': "fake_platform", 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'platform': "fake_platform", 'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'platform': "fake_platform", 'max_batch_size': 8, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}],'dynamic_batching': {}}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
+
+    def test_remote_yes_params_specified(self):
+        """
+        Test remote mode with model_config_parameters specified
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            triton_launch_mode: remote
+            run_config_search_max_instance_count: 16
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        max_batch_size: [1,4,16]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1,2]
+            """)
+
+        expected_configs = [
+            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
+            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
+            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 16},
+            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
+            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
+            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 16}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_remote_no_params_specified(self):
+        """
+        Test remote mode with no model_config_parameters specified
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            triton_launch_mode: remote
+            run_config_search_max_instance_count: 4
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 8, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}],'dynamic_batching': {}}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs)
+
+    def test_search_subparameter(self):
+        """
+        Test that if a subparameter is swept, that it will not overwrite other subparameters
+
+        Param2 should exist in all results despite the fact that param1 is the one being swept.
+        Also, gpu_execution_accelerator (a nested subproperty) should still exist despite a
+        sibling property (cpu_execution_accelerator) being overwritten
+        """
+
+        # yapf: disable
+        protobuf = """
+            max_batch_size: 4
+            instance_group [
+            {
+                kind: KIND_GPU
+                count: 1
+            }
+            ]
+            parameters [
+                {
+                    key: "param1"
+                    value: { string_value: "param1_value" }
+                },
+                {
+                    key: "param2"
+                    value: { string_value: "param2_value" }
+                }
+            ]
+            optimization { execution_accelerators {
+                cpu_execution_accelerator : [ {
+                    name : "fake_cpu_accelerator"
+                }]
+                gpu_execution_accelerator : [ {
+                    name : "fake_gpu_accelerator"
+                }]
+            }}
+            """
+
+        yaml_str = ("""
+            profile_models:
+                my-model:
+                    model_config_parameters:
+                        optimization:
+                            execution_accelerators:
+                                cpu_execution_accelerator:
+                                - name: "new_cpu_accelerator"
+                        parameters:
+                            param1:
+                                string_value: ["foo", "bar"]
+            """)
+
+        expected_configs = [
+            {
+                'optimization':
+                {
+                    'execution_accelerators':
+                    {
+                        'cpu_execution_accelerator': [{'name': 'new_cpu_accelerator'}],
+                        'gpu_execution_accelerator': [{'name': 'fake_gpu_accelerator'}]
+                    }
+                },
+                'parameters':
+                {
+                    'param1': {'string_value': 'foo'},
+                    'param2': {'string_value': 'param2_value'}
+                },
+                'max_batch_size': 4,
+                'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]
+            },
+            {
+                'optimization':
+                {
+                    'execution_accelerators':
+                    {
+                        'cpu_execution_accelerator': [{'name': 'new_cpu_accelerator'}],
+                        'gpu_execution_accelerator': [{'name': 'fake_gpu_accelerator'}]
+                    }
+                },
+                'parameters':
+                {
+                    'param1': {'string_value': 'bar'},
+                    'param2': {'string_value': 'param2_value'}
+                },
+                'max_batch_size': 4,
+                'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]
+            },
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
+
+    def test_search_dynamic_batching_subparameter(self):
+        """
+        Test that if dynamic batching was already on with subparameters, they will not be overwritten in automatic search
+
+        Normally automatic search just turns on dynamic_batching, but in this case dynamic_batching is already
+        on in the user's default model config with a value for max_queue_delay_microseconds. That value should remain
+        unchanged in all model configs generated by the manual search
+        """
+
+        # yapf: disable
+        protobuf = """
+            max_batch_size: 4
+            instance_group [
+            {
+                kind: KIND_GPU
+                count: 1
+            }
+            ]
+            dynamic_batching {
+                max_queue_delay_microseconds: 100
+            }
+            """
+
+        yaml_str = ("""
+            run_config_search_max_instance_count: 4
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = [
+            {'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': { 'max_queue_delay_microseconds': '100'}},
+            {'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': { 'max_queue_delay_microseconds': '100'}},
+            {'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': { 'max_queue_delay_microseconds': '100'}},
+            {'max_batch_size': 8, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}],'dynamic_batching': { 'max_queue_delay_microseconds': '100'}}
+        ]
+        # yapf: enable
+
+        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
+
+    def test_apply_value_to_dict(self):
+        """
+        Test different combinations of input and existing value types for apply_value_to_dict()
+        """
+        # Both input and existing are scalar value in a dict
+        existing_dict = {"a": 1, "b": 2}
+        expected_dict = {"a": 3, "b": 2}
+        BaseModelConfigGenerator._apply_value_to_dict("a", 3, existing_dict)
+        self.assertEqual(existing_dict, expected_dict)
+
+        # Input is scalar, existing is dict
+        existing_dict = {"a": 1, "b": {"c": 5, "d": 6}}
+        expected_dict = {"a": 1, "b": 2}
+        BaseModelConfigGenerator._apply_value_to_dict("b", 2, existing_dict)
+        self.assertEqual(existing_dict, expected_dict)
+
+        # Input is dict, existing is scalar
+        existing_dict = {"a": 1, "b": 3}
+        expected_dict = {"a": 1, "b": {"c": 7, "d": 8}}
+        BaseModelConfigGenerator._apply_value_to_dict(
+            "b", {"c": 7, "d": 8}, existing_dict
+        )
+        self.assertEqual(existing_dict, expected_dict)
+
+        # Input and dict are both dicts
+        existing_dict = {"a": 1, "b": {"c": {"e": 9, "f": 10}, "d": 6}}
+        expected_dict = {"a": 1, "b": {"c": {"e": 11, "f": 10}, "d": 6}}
+        BaseModelConfigGenerator._apply_value_to_dict(
+            "b", {"c": {"e": 11}}, existing_dict
+        )
+        self.assertEqual(existing_dict, expected_dict)
+
+    def test_early_exit_off_automatic_asserts(self):
+        """
+        Test that passing early_exit=False for automatic search raises an assert
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+
+        expected_configs = []
+        # yapf: enable
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self._run_and_test_model_config_generator(
+                yaml_str, expected_configs, early_exit_enable=False
+            )
+
+    def test_early_exit_on_automatic(self):
+        """
+        Test that automatic mode will early exit max_batch_size when throughput plateaus
+        """
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            run_config_search_max_instance_count: 3
+            run_config_search_max_model_batch_size: 8
+            """)
+
+        expected_configs = [
+            {'max_batch_size': 1, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 1, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 1, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+            {'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
+        ]
+
+        with patch.object(TestModelConfigGenerator,
+                          "_get_next_fake_throughput") as mock_method:
+            mock_method.side_effect = [
+                1, 2, 4, 8,  # 1 instance
+                1, 1,        # 2 instances. Yes backoff
+                1, 2, 4, 8   # 3 instances
+            ]
+            self._run_and_test_model_config_generator(yaml_str,
+                                                      expected_configs,
+                                                      early_exit_enable=True)
+        # yapf: enable
+
+    def test_early_exit_off_manual(self):
+        """
+        Test that manual mode will not early exit despite throughput plateauing despite because early_exit_enable=False
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        max_batch_size: [1,2,4,8]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1,2]
+            """)
+
+        expected_configs = [
+            {'max_batch_size': 1, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 1, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 4, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+        ]
+        # yapf: enable
+
+        with patch.object(
+            TestModelConfigGenerator, "_get_next_fake_throughput"
+        ) as mock_method:
+            mock_method.return_value = 1
+            self._run_and_test_model_config_generator(
+                yaml_str, expected_configs, early_exit_enable=False
+            )
+
+    def test_early_exit_on_manual(self):
+        """
+        Test that manual mode will early exit when throughput plateaus when early_exit_enable=True
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        max_batch_size: [1,2,3,4]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1,2,3]
+            """)
+
+        expected_configs = [
+            {'max_batch_size': 1, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 3, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 1, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 1, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 3, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+            {'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
+        ]
+
+        with patch.object(TestModelConfigGenerator,
+                          "_get_next_fake_throughput") as mock_method:
+            mock_method.side_effect = [
+                1, 2, 4, 8,  # 1 instance
+                1, 1,        # 2 instances. Yes backoff
+                1, 2, 4, 8   # 3 instances
+            ]
+            self._run_and_test_model_config_generator(yaml_str,
+                                                      expected_configs,
+                                                      early_exit_enable=True)
+        # yapf: enable
+
+    def test_extract_model_name_from_variant_name(self):
+        input_output_pairs = {}
+        input_output_pairs[
+            "onnx_int32_int32_int32_config_default"
+        ] = "onnx_int32_int32_int32"
+        input_output_pairs["onnx_int32_int32_int32_config_2"] = "onnx_int32_int32_int32"
+        input_output_pairs["onnx_int32_int32_int32"] = "onnx_int32_int32_int32"
+
+        for variant_name, expected_model_name in input_output_pairs.items():
+            model_name = BaseModelConfigGenerator.extract_model_name_from_variant_name(
+                variant_name
+            )
+            self.assertEqual(model_name, expected_model_name)
+
+    def _run_and_test_model_config_generator(
+        self,
+        yaml_str,
+        expected_configs,
+        protobuf="max_batch_size: 8",
+        default_only=False,
+        early_exit_enable=True,
+    ):
+        """
+        Main function that creates a config from the yaml_str, runs it through
+        ModelConfigGenerator, and compares the resulting model_configs vs the expected_configs
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+
+        # Use mock model config or else TritonModelAnalyzerException will be thrown as it tries to read from disk
+        self.mock_model_config = MockModelConfig(protobuf)
+        self.mock_model_config.start()
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        profile_models = []
+        for model in config.profile_models:
+            profile_models.append(
+                ModelProfileSpec(model, config, MagicMock(), MagicMock())
+            )
+
+        # Fake out a client that can return a 'model_config' dict with
+        # a valid name (only used by remote mode)
+        #
+        fake_client = MagicMock()
+        fake_client.get_model_config = lambda name, retry_count: {"name": name}
+
+        try:
+            mcg = ModelConfigGeneratorFactory.create_model_config_generator(
+                config,
+                MagicMock(),
+                profile_models[0],
+                fake_client,
+                ModelVariantNameManager(),
+                default_only=default_only,
+                early_exit_enable=early_exit_enable,
+            )
+
+            model_configs = []
+            for model_config_variant in mcg.get_configs():
+                mcg.set_last_results(self._get_next_fake_results())
+                model_config_dict = model_config_variant.model_config.get_config()
+                model_configs.append(model_config_dict)
+
+            self.assertEqual(len(expected_configs), len(model_configs))
+
+            # Rip out the model name (so the order doesn't have to exactly match),
+            # but verify that it exists and is not none
+            #
+            for config in model_configs:
+                name = config.pop("name", None)
+                self.assertIsNotNone(name)
+
+            # Confirm the configs match
+            #
+            for config in expected_configs:
+                self.assertIn(config, model_configs)
+
+        finally:
+            self.mock_model_config.stop()
+
+    def _get_next_fake_results(self):
+        throughput = self._get_next_fake_throughput()
+
+        measurement = construct_run_config_measurement(
+            model_name=MagicMock(),
+            model_config_names=["test_model_config_name"],
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=MagicMock(),
+            non_gpu_metric_values=[{"perf_throughput": throughput}],
+        )
+
+        return [measurement]
+
+    def _get_next_fake_throughput(self):
+        self._fake_throughput += 1
+        return self._fake_throughput
+
+    def setUp(self):
+        # Mock path validation
+        self.mock_os = MockOSMethods(
+            mock_paths=["model_analyzer.config.input.config_utils"]
+        )
+        self.mock_os.start()
+
+    def tearDown(self):
+        self.mock_os.stop()
+        patch.stopall()
+        ModelConfig._default_config_dict = {}
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_config_measurement.py b/src/c++/perf_analyzer/tests/test_model_config_measurement.py
new file mode 100755
index 000000000..e760f1efa
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_model_config_measurement.py
@@ -0,0 +1,242 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import json
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.result.model_config_measurement import ModelConfigMeasurement
+from tests.common.test_utils import convert_non_gpu_metrics_to_data, default_encode
+
+from .common import test_result_collector as trc
+
+
+class TestModelConfigMeasurement(trc.TestResultCollector):
+    def setUp(self):
+        self.model_config_name = "modelA"
+        self.model_specific_pa_params = {"batch_size": 1, "concurrency-range": 1}
+
+        self.non_gpu_metric_values = {
+            "perf_throughput": 1000,
+            "perf_latency_p99": 20,
+            "cpu_used_ram": 1000,
+        }
+
+        self.mcmA = self._construct_model_config_measurement(
+            self.model_config_name,
+            self.model_specific_pa_params,
+            self.non_gpu_metric_values,
+        )
+
+        mcmB_non_gpu_metric_values = {
+            "perf_throughput": 2000,
+            "perf_latency_p99": 40,
+            "cpu_used_ram": 1000,
+        }
+
+        self.mcmB = self._construct_model_config_measurement(
+            "modelB", self.model_specific_pa_params, mcmB_non_gpu_metric_values
+        )
+
+        self.mcmC = self._construct_model_config_measurement(
+            "modelC", self.model_specific_pa_params, {}
+        )
+
+        self.mcmD = self._construct_model_config_measurement(
+            "modelD", self.model_specific_pa_params, {}
+        )
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_init(self):
+        """
+        Test that values are properly initialized
+        """
+        self.assertEqual(self.mcmA.model_config_name(), self.model_config_name)
+        self.assertEqual(
+            self.mcmA.model_specific_pa_params(), self.model_specific_pa_params
+        )
+        self.assertEqual(
+            self.mcmA.non_gpu_data(),
+            convert_non_gpu_metrics_to_data(self.non_gpu_metric_values),
+        )
+
+    def test_get_metric_found(self):
+        """
+        Test that non-gpu metrics can be correctly returned
+        """
+        non_gpu_data = convert_non_gpu_metrics_to_data(self.non_gpu_metric_values)
+
+        self.assertEqual(self.mcmA.get_metric("perf_throughput"), non_gpu_data[0])
+        self.assertEqual(self.mcmA.get_metric("perf_latency_p99"), non_gpu_data[1])
+        self.assertEqual(self.mcmA.get_metric("cpu_used_ram"), non_gpu_data[2])
+
+    def test_get_metric_not_found(self):
+        """
+        Test that an incorrect metric search returns None
+        """
+        self.assertEqual(self.mcmA.get_metric("XXXXX"), None)
+
+    def test_get_metric_value_found(self):
+        """
+        Test that non-gpu metric values can be correctly returned
+        """
+        self.assertEqual(
+            self.mcmA.get_metric_value("perf_throughput"),
+            self.non_gpu_metric_values["perf_throughput"],
+        )
+        self.assertEqual(
+            self.mcmA.get_metric_value("perf_latency_p99"),
+            self.non_gpu_metric_values["perf_latency_p99"],
+        )
+        self.assertEqual(
+            self.mcmA.get_metric_value("cpu_used_ram"),
+            self.non_gpu_metric_values["cpu_used_ram"],
+        )
+
+    def test_get_metric_value_not_found(self):
+        """
+        Test that an incorrect metric value search returns the correct value
+        """
+        self.assertEqual(self.mcmA.get_metric_value("XXXXX"), 0)
+        self.assertEqual(self.mcmA.get_metric_value("XXXXX", 100), 100)
+
+    def test_set_metric_weighting(self):
+        # Default
+        self.assertEqual({"perf_throughput": 1}, self.mcmA._metric_weights)
+
+        self.mcmA.set_metric_weighting({"perf_throughput": 2, "perf_latency_p99": 3})
+        expected_mw = {"perf_throughput": 2 / 5, "perf_latency_p99": 3 / 5}
+        self.assertEqual(expected_mw, self.mcmA._metric_weights)
+
+    def test_calculate_weighted_percentage_gain(self):
+        """
+        Test that weighted percentages are returned correctly
+        """
+        self.mcmA.set_metric_weighting({"perf_throughput": 1})
+        self.mcmB.set_metric_weighting({"perf_throughput": 1})
+
+        # throuhput: mcmA: 1000, mcmB: 2000
+        self.assertEqual(self.mcmA.calculate_weighted_percentage_gain(self.mcmB), -50)
+        self.assertEqual(self.mcmB.calculate_weighted_percentage_gain(self.mcmA), 100)
+
+        self.mcmA.set_metric_weighting({"perf_latency_p99": 1})
+        self.mcmB.set_metric_weighting({"perf_latency_p99": 1})
+
+        # latency: mcmA: 20, mcmB: 40
+        self.assertEqual(self.mcmA.calculate_weighted_percentage_gain(self.mcmB), 100)
+        self.assertEqual(self.mcmB.calculate_weighted_percentage_gain(self.mcmA), -50)
+
+        # This illustrates why we need to use score, not percentages to determine
+        # which model is better. In both cases we will (correctly) report that
+        # mcmA/B is 25% better than the other, even though they are equal
+        #
+        # mcmA has 50% worse throughput, but 100% better latency
+        # mcmB has 100% better latency, but 50% worse throughput
+        self.mcmA.set_metric_weighting({"perf_throughput": 1, "perf_latency_p99": 1})
+        self.mcmB.set_metric_weighting({"perf_throughput": 1, "perf_latency_p99": 1})
+        self.assertEqual(self.mcmA, self.mcmB)
+        self.assertEqual(self.mcmA.calculate_weighted_percentage_gain(self.mcmB), 25)
+        self.assertEqual(self.mcmB.calculate_weighted_percentage_gain(self.mcmA), 25)
+
+    def test_is_better_than(self):
+        """
+        Test that individual metric comparison works as expected
+        """
+        self.mcmA.set_metric_weighting({"perf_throughput": 1})
+
+        # throughput: 1000 is not better than 2000
+        self.assertFalse(self.mcmA.is_better_than(self.mcmB))
+        self.assertLess(self.mcmA, self.mcmB)
+
+        self.mcmA.set_metric_weighting({"perf_latency_p99": 1})
+
+        # latency: 20 is better than 40
+        self.assertTrue(self.mcmA.is_better_than(self.mcmB))
+        self.assertGreater(self.mcmA, self.mcmB)
+
+    def test_is_better_than_combo(self):
+        """
+        Test that combination metric comparison works as expected
+        """
+        # throuhput: 1000 vs. 2000 (worse), latency: 20 vs. 40 (better)
+        # with latency bias mcmA is better
+        self.mcmA.set_metric_weighting({"perf_throughput": 1, "perf_latency_p99": 3})
+
+        self.assertTrue(self.mcmA.is_better_than(self.mcmB))
+
+    def test_is_better_than_empty(self):
+        """
+        Test for correct return values when comparing for/against an empty set
+        """
+        self.mcmA.set_metric_weighting({"perf_throughput": 1})
+        self.mcmC.set_metric_weighting({"perf_throughput": 1})
+
+        self.assertTrue(self.mcmA.is_better_than(self.mcmC))
+        self.assertFalse(self.mcmC.is_better_than(self.mcmA))
+        self.assertEqual(self.mcmC, self.mcmD)
+
+    def test__eq__(self):
+        """
+        Test that individual metric equality works as expected
+        """
+        self.mcmA.set_metric_weighting({"cpu_used_ram": 10})
+
+        self.assertEqual(self.mcmA, self.mcmB)
+
+    def test__eq__combo(self):
+        """
+        Test that combination metric equality works as expected
+        """
+        # throuhput: 1000 vs. 2000 (worse), latency: 20 vs. 40 (better)
+        # with no bias they are equal
+        self.mcmA.set_metric_weighting({"perf_throughput": 1, "perf_latency_p99": 1})
+
+        self.assertEqual(self.mcmA, self.mcmB)
+
+    def test_from_dict(self):
+        """
+        Test to ensure class can be correctly restored from a dictionary
+        """
+        mcmA_json = json.dumps(self.mcmA, default=default_encode)
+
+        mcmA_from_dict = ModelConfigMeasurement.from_dict(json.loads(mcmA_json))
+
+        self.assertEqual(
+            mcmA_from_dict.model_config_name(), self.mcmA.model_config_name()
+        )
+        self.assertEqual(
+            mcmA_from_dict.model_specific_pa_params(),
+            self.mcmA.model_specific_pa_params(),
+        )
+        self.assertEqual(mcmA_from_dict.non_gpu_data(), self.mcmA.non_gpu_data())
+
+        # Catchall in case something new is added
+        self.assertEqual(mcmA_from_dict, self.mcmA)
+
+    def _construct_model_config_measurement(
+        self, model_config_name, model_specific_pa_params, non_gpu_metric_values
+    ):
+        non_gpu_data = convert_non_gpu_metrics_to_data(non_gpu_metric_values)
+
+        return ModelConfigMeasurement(
+            model_config_name, model_specific_pa_params, non_gpu_data
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_constraints.py b/src/c++/perf_analyzer/tests/test_model_constraints.py
new file mode 100755
index 000000000..a265f3bdb
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_model_constraints.py
@@ -0,0 +1,119 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.result.model_constraints import ModelConstraints
+
+from .common import test_result_collector as trc
+
+
+class TestModelConstraints(trc.TestResultCollector):
+    def tearDown(self):
+        patch.stopall()
+
+    def test_to_dict(self):
+        """
+        Test that constraints object to dict
+        """
+        constraints_dict_1 = {
+            "perf_latency_p99": {"max": 100},
+            "gpu_used_memory": {"max": 20000},
+        }
+        constraints_dict_2 = {}
+
+        constraints_obj_1 = ModelConstraints(constraints=constraints_dict_1)
+        constraints_obj_2 = ModelConstraints(constraints=constraints_dict_2)
+
+        self.assertEqual(constraints_obj_1.to_dict(), constraints_dict_1)
+        self.assertEqual(constraints_obj_2.to_dict(), constraints_dict_2)
+
+    def test_has_metric(self):
+        """
+        Test that constraints object has a given key
+        """
+        constraints_dict_1 = {
+            "perf_latency_p99": {"max": 100},
+            "gpu_used_memory": {"max": 20000},
+        }
+        constraints_dict_2 = {}
+
+        constraints_obj_1 = ModelConstraints(constraints=constraints_dict_1)
+        constraints_obj_2 = ModelConstraints(constraints=constraints_dict_2)
+
+        self.assertTrue(constraints_obj_1.has_metric("perf_latency_p99"))
+        self.assertTrue(constraints_obj_1.has_metric("gpu_used_memory"))
+        self.assertFalse(constraints_obj_1.has_metric("perf_throughput"))
+        self.assertFalse(constraints_obj_1.has_metric(""))
+        self.assertFalse(constraints_obj_2.has_metric("perf_latency_p99"))
+        self.assertFalse(constraints_obj_2.has_metric("gpu_used_memory"))
+
+    def test_getitem(self):
+        """
+        Test that metrics are subscriptable from constraints object
+        """
+        constraints_dict = {
+            "perf_latency_p99": {"max": 100},
+            "gpu_used_memory": {"max": 20000},
+        }
+
+        constraints_obj = ModelConstraints(constraints=constraints_dict)
+
+        self.assertEqual(constraints_obj["perf_latency_p99"], {"max": 100})
+        self.assertEqual(constraints_obj["gpu_used_memory"], {"max": 20000})
+
+    def test_bool(self):
+        """
+        Test that constraints are empty
+        """
+        constraints_dict_1 = {
+            "perf_latency_p99": {"max": 100},
+            "gpu_used_memory": {"max": 20000},
+        }
+        constraints_dict_2 = {}
+
+        constraints_obj_1 = ModelConstraints(constraints=constraints_dict_1)
+        constraints_obj_2 = ModelConstraints(constraints=constraints_dict_2)
+
+        self.assertTrue(constraints_obj_1)
+        self.assertFalse(constraints_obj_2)
+
+    def test_eq(self):
+        """
+        Test that constraint objects are equal
+        """
+        constraints_dict_1 = {
+            "perf_latency_p99": {"max": 100},
+            "gpu_used_memory": {"max": 20000},
+        }
+        constraints_dict_2 = {
+            "perf_latency_p99": {"max": 100},
+            "gpu_used_memory": {"max": 20000},
+        }
+        constraints_dict_3 = {}
+
+        constraints_obj_1 = ModelConstraints(constraints=constraints_dict_1)
+        constraints_obj_2 = ModelConstraints(constraints=constraints_dict_2)
+        constraints_obj_3 = ModelConstraints(constraints=constraints_dict_3)
+
+        self.assertEqual(constraints_obj_1, constraints_obj_2)
+        self.assertNotEqual(constraints_obj_3, constraints_obj_2)
+        self.assertNotEqual(constraints_obj_1, constraints_obj_3)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_manager.py b/src/c++/perf_analyzer/tests/test_model_manager.py
new file mode 100755
index 000000000..0370c6c77
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_model_manager.py
@@ -0,0 +1,1319 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.input.objects.config_model_profile_spec import (
+    ConfigModelProfileSpec,
+)
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.model_manager import ModelManager
+from model_analyzer.record.metrics_manager import MetricsManager
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+from model_analyzer.triton.model.model_config import ModelConfig
+
+from .common import test_result_collector as trc
+from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
+from .mocks.mock_model_config import MockModelConfig
+from .mocks.mock_run_configs import MockRunConfigs
+
+
+class MetricsManagerSubclass(MetricsManager):
+    """
+    Overrides execute_model_run_config() to gather a list of MockRunConfigs that
+    contain the configured values of each would-be 'executed' run_config
+    """
+
+    def __init__(self, config, client, server, gpus, result_manager, state_manager):
+        super().__init__(config, client, server, gpus, result_manager, state_manager)
+        self._configs = MockRunConfigs()
+        self._perf_throughput = 1
+
+    def get_run_configs(self):
+        """Return the list of configs that would have been 'executed'"""
+        return self._configs
+
+    def execute_run_config(self, config):
+        self._configs.add_from_model_run_config(config.model_run_configs()[0])
+        return self._get_next_measurements()
+
+    def _get_next_measurements(self):
+        """Return fake measurements as if the run_configs had been executed"""
+
+        throughput_value = self._get_next_perf_throughput_value()
+        if throughput_value is None:
+            return None
+        else:
+            return construct_run_config_measurement(
+                model_name=MagicMock(),
+                model_config_names=["test_model_config_name"],
+                model_specific_pa_params=MagicMock(),
+                gpu_metric_values=MagicMock(),
+                non_gpu_metric_values=[{"perf_throughput": throughput_value}],
+            )
+
+    def _get_next_perf_throughput_value(self):
+        self._perf_throughput *= 2
+        return self._perf_throughput
+
+
+class TestModelManager(trc.TestResultCollector):
+    def tearDown(self):
+        patch.stopall()
+        ModelConfig._default_config_dict = {}
+
+    def __init__(self, methodname):
+        super().__init__(methodname)
+        self._args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+
+        self._model_config_protobuf = """
+            name: "test_model"
+            platform: "tensorflow_graphdef"
+            max_batch_size: 8
+            input [
+            {
+                name: "INPUT__0"
+                data_type: TYPE_FP32
+                dims: [16]
+            }
+            ]
+            output [
+            {
+                name: "OUTPUT__0"
+                data_type: TYPE_FP32
+                dims: [16]
+            }
+            ]
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 1
+            }
+            ]
+            """
+
+    def test_full_sweep(self):
+        """
+        Test a normal full sweep of options
+        """
+        expected_ranges = [
+            {
+                "instances": [1, 2, 3, 4, 5],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1],
+                "max_batch_size": [1, 2, 4, 8],
+                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 128
+            run_config_search_max_instance_count: 5
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_another_full_sweep(self):
+        """
+        Test another full sweep of options
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1, 2, 3, 4, 5, 6, 7],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8, 16, 32],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8, 16, 32],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 32
+            run_config_search_max_instance_count: 7
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_run_search_disable(self):
+        """
+        Test with run_config_search_disable=True
+
+        Expect 1 result that matches the default configuration because no manual
+        search options provided and automatic search disabled/ignored
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 32
+            run_config_search_max_instance_count: 7
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: True
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_manual_concurrency(self):
+        """
+        Test with manually specified concurrencies
+        """
+        expected_ranges = [
+            {
+                "instances": [1, 2, 3, 4, 5, 6, 7],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [5, 7],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [5, 7],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 32
+            run_config_search_max_instance_count: 7
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            concurrency: [5, 7]
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_remote_mode(self):
+        """
+        Test remote mode
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1, 2],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8, 16],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8, 16],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 16
+            run_config_search_max_instance_count: 2
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            triton_launch_mode: remote
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_manual_parameters(self):
+        """
+        Test with manually specified concurrencies and batch sizes
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1, 2, 3, 4, 5, 6, 7],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1, 2, 3],
+                "max_batch_size": [8],
+                "concurrency": [2, 10, 18, 26, 34, 42, 50, 58],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1, 2, 3],
+                "max_batch_size": [8],
+                "concurrency": [2, 10, 18, 26, 34, 42, 50, 58],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 512
+            run_config_search_max_instance_count: 7
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            concurrency:
+                start: 2
+                stop: 64
+                step: 8
+            batch_sizes: 1,2,3
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_no_early_exit_client_batch_size(self):
+        """
+        Test that we will NOT early exit client batch size by default
+        """
+        self._test_early_exit_client_batch_size_helper(early_exit=False)
+
+    def test_early_exit_client_batch_size(self):
+        """
+        Test that we will early exit client batch size if
+        --early-exit-enable is true
+        """
+        self._test_early_exit_client_batch_size_helper(early_exit=True)
+
+    def _test_early_exit_client_batch_size_helper(self, early_exit):
+        args = self._args.copy()
+
+        if early_exit:
+            args.append("--early-exit-enable")
+
+            expected_ranges = [
+                {
+                    "instances": [1],
+                    "kind": ["KIND_GPU"],
+                    "batching": [0],
+                    "batch_sizes": [1, 2, 3, 4],
+                    "max_batch_size": [8],
+                    "concurrency": [1, 2, 4, 8, 16],
+                },
+                {
+                    "instances": [1],
+                    "kind": ["KIND_CPU"],
+                    "batching": [None],
+                    "batch_sizes": [1, 2, 3, 4, 7],
+                    "max_batch_size": [8],
+                    "concurrency": [1, 2, 4, 8],
+                },
+            ]
+
+        else:
+            expected_ranges = [
+                {
+                    "instances": [1],
+                    "kind": ["KIND_GPU"],
+                    "batching": [0],
+                    "batch_sizes": [1, 2, 3, 4, 7],
+                    "max_batch_size": [8],
+                    "concurrency": [1, 2, 4, 8, 16],
+                },
+                {
+                    "instances": [1],
+                    "kind": ["KIND_CPU"],
+                    "batching": [None],
+                    "batch_sizes": [1, 2, 3, 4, 7],
+                    "max_batch_size": [8],
+                    "concurrency": [1, 2, 4, 8, 16],
+                },
+            ]
+
+        yaml_content = """
+            profile_models: test_model
+            run_config_search_max_instance_count: 1
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            concurrency: 16,1,2,4,8
+            batch_sizes: 3,7,1,4,2
+            """
+
+        with patch.object(
+            MetricsManagerSubclass, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            # yapf: disable
+            side_effect = [
+                # Default config, bs=1, concurrency 1,2,4,8
+                # Will early exit for concurrency
+                # "Best" result for bs early exit is 5
+                5, 5, 5, 5,
+
+                # Default config, bs=2, concurrency 1,2,4,8
+                # Will early exit for concurrency
+                # "Best" result for bs early exit is 4
+                4, 4, 4, 4,
+
+                # Default config, bs=3, concurrency 1,2,4,8
+                # Will early exit for concurrency
+                # "Best" result for bs early exit is 6
+                6, 6, 6, 6,
+
+                # Default config, bs=4, concurrency 1,2,4,8
+                # Will early exit for concurrency
+                # "Best" result for bs early exit is 5
+                # We will not early exit batch size
+                5, 5, 5, 5,
+
+                # Default config, bs=7, concurrency 1,2,4,8
+                # Will not early exit for concurrency
+                # "Best" result for bs early exit is 1
+                # We are done sweeping batch size
+                1, 1, 1, 1,
+
+
+                # 1 instance, bs=1, concurrency 1,2,4,8,16
+                # Will not early exit for concurrency
+                # "Best" result for bs early exit is 10
+                1, 1, 10, 1, 1,
+
+                # 1 instance, bs=2, concurrency 1,2,4,8,16
+                # Will not early exit for concurrency
+                # "Best" result for bs early exit is 9
+                1, 9, 1, 1, 2,
+
+                # 1 instance, bs=3, concurrency 1,2,4,8,16
+                # Will not early exit for concurrency
+                # "Best" result for bs early exit is 8
+                1, 1, 1, 8, 3,
+
+                # 1 instance, bs=4, concurrency 1,2,4,8,16
+                # Will not early exit for concurrency
+                # "Best" result for bs early exit is 7
+                # Will early exit batch size now
+                1, 1, 7, 1, 4
+            ]
+            # Add a bunch of extra results for the no-early-exit case
+            side_effect.extend([1]*100)
+            mock_method.side_effect = side_effect
+            # yapf: enable
+
+            self._test_model_manager(yaml_content, expected_ranges, args=args)
+
+    def test_model_config_parameters(self):
+        """
+        Test with manually specified model config parameters
+
+        In this case we don't automatically search instances or dynamic_batching
+        since model config parameters are specified.
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batch_sizes": [1],
+                "max_batch_size": [1, 3, 7, 8, 13],
+                "concurrency": [1, 2, 4, 8],
+            }
+        ]
+
+        yaml_str = """
+            run_config_search_max_concurrency: 8
+            run_config_search_max_instance_count: 16
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        max_batch_size: [1,3,7,13]
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_default_config_always_run_no_dynamic_batching_off(self):
+        """
+        Test that the default config is run even when manual search excludes that case
+        In this case, default config is (1 instance, max_batch_size=8, dynamic batching off)
+        We should have a case of dynamic_batching off even though manual search only has it on
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [0],
+                "max_queue_delay": ["200", "300"],
+                "batch_sizes": [1],
+                "max_batch_size": [1, 2, 4, 8, 16],
+                "concurrency": [1, 2, 4, 8],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8],
+            },
+        ]
+
+        yaml_str = """
+            run_config_search_max_concurrency: 8
+            run_config_search_max_instance_count: 16
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        max_batch_size: [1,2,4,8,16]
+                        dynamic_batching:
+                            max_queue_delay_microseconds: [200, 300]
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_default_config_always_run_wrong_instances(self):
+        """
+        Test that the default config is run even when manual search excludes that case
+        In this case, default config is (2 instances, max_batch_size=8, dynamic batching off)
+        We should have a 2-instance case even though manual search only has 1-instance
+        """
+
+        self._model_config_protobuf = """
+            name: "test_model"
+            platform: "tensorflow_graphdef"
+            max_batch_size: 8
+            input [
+            {
+                name: "INPUT__0"
+                data_type: TYPE_FP32
+                dims: [16]
+            }
+            ]
+            output [
+            {
+                name: "OUTPUT__0"
+                data_type: TYPE_FP32
+                dims: [16]
+            }
+            ]
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 2
+            }
+            ]
+            """
+
+        expected_ranges = [
+            {
+                "instances": [1],
+                "kind": ["KIND_GPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4],
+            },
+            {
+                "instances": [2],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4],
+            },
+        ]
+
+        yaml_str = """
+            run_config_search_max_concurrency: 4
+            run_config_search_max_instance_count: 16
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: 1
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_default_config_always_run_cpu_vs_gpu(self):
+        """
+        If the default configuration had KIND_CPU, make sure it is run (even if everything
+        else is the same)
+        """
+
+        self._model_config_protobuf = """
+            name: "test_model"
+            platform: "tensorflow_graphdef"
+            max_batch_size: 8
+            input [
+            {
+                name: "INPUT__0"
+                data_type: TYPE_FP32
+                dims: [16]
+            }
+            ]
+            output [
+            {
+                name: "OUTPUT__0"
+                data_type: TYPE_FP32
+                dims: [16]
+            }
+            ]
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 1
+            }
+            ]
+            """
+
+        expected_ranges = [
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_GPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4],
+            },
+        ]
+
+        yaml_str = """
+            run_config_search_max_concurrency: 4
+            run_config_search_max_instance_count: 16
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: 1
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_default_config_always_run_automatic_search(self):
+        """
+        Test that the default config is run even when automatic search excludes that case
+        In this case, default config is (4 instance, CPU, max_batch_size=8, dynamic batching off)
+        We should have this 4 instance case though run_config_search_max_instance_count=1
+        """
+
+        self._model_config_protobuf = """
+            name: "test_model"
+            platform: "tensorflow_graphdef"
+            max_batch_size: 8
+            input [
+            {
+                name: "INPUT__0"
+                data_type: TYPE_FP32
+                dims: [16]
+            }
+            ]
+            output [
+            {
+                name: "OUTPUT__0"
+                data_type: TYPE_FP32
+                dims: [16]
+            }
+            ]
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 4
+            }
+            ]
+            """
+
+        expected_ranges = [
+            {
+                "instances": [1],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4],
+            },
+            {
+                "instances": [4],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4],
+            },
+        ]
+
+        yaml_str = """
+            run_config_search_max_concurrency: 4
+            run_config_search_max_instance_count: 1
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            profile_models: test_model
+            """
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_throughput_early_exit_minimum_runs(self):
+        """
+        Test that there is an early back off when sweeping concurrency
+
+        The behavior is that MA will try at least 4 concurrencies. If
+        at that point none of the last 3 attempts have had satisfactory
+        gain, it will stop
+
+        This test hardcodes the 'throughput' to 1, so for all model
+        configs the gain will be invalid and it will only try 4
+        concurrencies of (1,2,4,8) despite max_concurrency=128
+
+        This test also has multiple batch sizes to make sure that hitting
+        an early exit on one doesn't cause the next one to be skipped
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1, 2],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1, 2],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1, 2],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 128
+            run_config_search_max_instance_count: 2
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            batch_sizes: 1,2
+            """
+
+        with patch.object(
+            MetricsManagerSubclass, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.return_value = 1
+            self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_no_early_exit_if_not_auto_search(self):
+        """
+        Test that there is NOT an early back off when sweeping concurrency if not in auto sweep mode
+
+        This test hardcodes the 'throughput' to 1, so for all model
+        configs the gain will be invalid. However, it should still sweep
+        to concurrency of 128 due to the fact that it was manually specified instead
+        of an auto-search
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1, 2],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1, 2],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1, 2],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 128
+            run_config_search_max_instance_count: 2
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            batch_sizes: 1,2
+            concurrency: 1,2,4,8,16,32,64,128
+            """
+
+        with patch.object(
+            MetricsManagerSubclass, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.return_value = 1
+            self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_throughput_early_exit(self):
+        """
+        Test that there is an early back off when sweeping concurrency
+
+        The behavior is that MA stop if it had 4 concurrencies in a row
+        without any valid gain amongst any of them
+
+        This test sets the 'throughput' to [1,2,4,8,16,16,16,16], which
+        will cause an early exit after trying the 8th concurrency (128)
+        instead of searching all the way to 2048
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 2048
+            run_config_search_max_instance_count: 1
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            """
+
+        with patch.object(
+            MetricsManagerSubclass, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = [
+                1,
+                2,
+                4,
+                8,
+                16,
+                16,
+                16,
+                16,
+                1,
+                2,
+                4,
+                8,
+                16,
+                16,
+                16,
+                16,
+            ]
+            self._test_model_manager(yaml_str, expected_ranges)
+
+    @patch("model_analyzer.model_manager.INVALID_MEASUREMENT_THRESHOLD", 999)
+    def test_bad_result_early_PA_exit(self):
+        """
+        Test that there is an early back off for bad result (out of memory)
+
+        If no measurements are returned in an attempt, no further concurrencies
+        should be tried.
+
+        This test hardcodes the measurements to be empty (bad result), so for all
+        model configs it will only try 1 concurrency despite max_concurrency=128
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1, 2],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 128
+            run_config_search_max_instance_count: 2
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            """
+
+        with patch.object(
+            MetricsManagerSubclass, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.return_value = None
+            self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_report_failure_no_measurements(self):
+        """
+        Test that MA takes an exception if we detect no measurements returned from
+        PA at the start of profile
+        """
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 128
+            run_config_search_max_instance_count: 2
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 8
+            run_config_search_disable: False
+            """
+
+        with patch.object(
+            MetricsManagerSubclass, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.return_value = None
+            with self.assertRaises(TritonModelAnalyzerException):
+                self._test_model_manager(yaml_str, None)
+
+    def test_lower_throughput_early_batch_size_exit(self):
+        """
+        Test that there is an early back off for throughput decreasing
+        when sweeping max_batch_size
+
+        If a list of measurements is provided with a lower max throughput than the previous
+        list of measurements, then we should early exit that max_batch_size sweep
+
+        The test is set up such that at instance count 1, going from max_batch_size of 16 to 32
+        does not increase the throughput, and thus we should not continue stepping max_batch_size.
+        The same is true for instance count 2, going from max_batch_size of 32 to 64
+        """
+
+        expected_ranges = [
+            # Instance count of 1 will stop after max_batch_size=32
+            {
+                "instances": [1],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1],
+                "max_batch_size": [8, 16, 32],
+                "concurrency": [1, 2, 4],
+            },
+            # Instance count of 2 will stop after max_batch_size=64
+            {
+                "instances": [2],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1],
+                "max_batch_size": [8, 16, 32, 64],
+                "concurrency": [1, 2, 4],
+            },
+            # Default config
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 4
+            run_config_search_max_instance_count: 2
+            run_config_search_min_model_batch_size: 8
+            run_config_search_disable: False
+            """
+
+        with patch.object(
+            MetricsManagerSubclass, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            # yapf: disable
+            mock_method.side_effect = [
+                1, 2, 4,     # Default config, concurrency 1,2,4
+                1, 2, 4,     # 1 Instance, Batch size 8, concurrency 1,2,4
+                2, 4, 8,     # 1 Instance, Batch size 16, concurrency 1,2,4
+                2, 4, 8,     # 1 Instance, Batch size 32, concurrency 1,2,4
+                1, 2, 4,     # 1 Instance, Batch size 8, concurrency 1,2,4
+                8, 4, 2,     # 1 Instance, Batch size 16, concurrency 1,2,4
+                4, 8, 16,    # 1 Instance, Batch size 32, concurrency 1,2,4
+                4, 8, 16     # 1 Instance, Batch size 64, concurrency 1,2,4
+            ]
+            # yapf: enable
+
+            mock_method.return_value = None
+            self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_no_max_batch_size_sweep_if_protobuf_0(self):
+        """
+        Test that if the max_batch_size is 0 in the original model config,
+        then do not sweep max batch size or dynamic batching
+        """
+
+        self._model_config_protobuf = """
+            name: "test_model"
+            platform: "tensorflow_graphdef"
+            max_batch_size: 0
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 1
+            }
+            ]
+            """
+
+        expected_ranges = [
+            {
+                "instances": [1],
+                "kind": ["KIND_GPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "concurrency": [1, 2, 4],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1],
+                "concurrency": [1, 2, 4],
+            },
+        ]
+
+        yaml_str = """
+            run_config_search_max_concurrency: 4
+            run_config_search_max_instance_count: 1
+            run_config_search_disable: False
+            profile_models: test_model
+            """
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    def test_client_batch_size_never_above_max_batch_size(self):
+        """
+        When we are sweeping through values of PA batch size and model max_batch_size,
+        it should never be the case that PA batch size > model max_batch_size
+        """
+
+        expected_ranges = [
+            {
+                "instances": [1],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1, 2, 4, 8],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1, 2, 4, 8, 16],
+                "max_batch_size": [16],
+                "concurrency": [1, 2, 4, 8],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_GPU"],
+                "batching": [0],
+                "batch_sizes": [1, 2, 4, 8, 16, 32],
+                "max_batch_size": [32],
+                "concurrency": [1, 2, 4, 8],
+            },
+            {
+                "instances": [1],
+                "kind": ["KIND_CPU"],
+                "batching": [None],
+                "batch_sizes": [1, 2, 4, 8],
+                "max_batch_size": [8],
+                "concurrency": [1, 2, 4, 8],
+            },
+        ]
+
+        yaml_str = """
+            profile_models: test_model
+            run_config_search_max_concurrency: 8
+            run_config_search_max_instance_count: 1
+            run_config_search_min_model_batch_size: 8
+            run_config_search_max_model_batch_size: 32
+            run_config_search_disable: False
+            batch_sizes: 1,2,4,8,16,32,64
+            """
+
+        self._test_model_manager(yaml_str, expected_ranges)
+
+    @patch(
+        "model_analyzer.triton.model.model_config.ModelConfig.is_ensemble",
+        return_value=True,
+    )
+    def test_ensemble_illegal_checks(self, *args):
+        """
+        Test that RCS mode isn't set to brute and that multiple models are
+        not provided when profiling an ensemble model
+        """
+        yaml_str = """
+                  profile_models: ensemble_model, test_model
+                  """
+
+        args = self._args.copy()
+        args.append("--run-config-search-mode")
+        args.append("brute")
+
+        self.mock_model_config = MockModelConfig(self._model_config_protobuf)
+        self.mock_model_config.start()
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        state_manager = AnalyzerStateManager(config, MagicMock())
+        metrics_manager = MetricsManagerSubclass(
+            config, MagicMock(), MagicMock(), MagicMock(), MagicMock(), state_manager
+        )
+        model_manager = ModelManager(
+            config,
+            MagicMock(),
+            MagicMock(),
+            MagicMock(),
+            metrics_manager,
+            MagicMock(),
+            state_manager,
+            MagicMock(),
+        )
+
+        # Multiple model check
+        models = [
+            ConfigModelProfileSpec("ensemble_model"),
+            ConfigModelProfileSpec("test_model"),
+        ]
+        with self.assertRaises(TritonModelAnalyzerException):
+            model_manager._check_for_ensemble_model_incompatibility(models)
+
+        # RunConfigSearch check
+        models = [
+            ConfigModelProfileSpec("ensemble_model"),
+        ]
+        with self.assertRaises(TritonModelAnalyzerException):
+            model_manager._check_for_ensemble_model_incompatibility(models)
+        self.mock_model_config.stop()
+
+    @patch(
+        "model_analyzer.triton.model.model_config.ModelConfig.is_ensemble",
+        return_value=True,
+    )
+    def test_ensemble_makes_quick_default(self, *args):
+        """
+        Test that the default RCS mode is switched from brute to quick
+        when profiling an ensemble model
+        """
+        yaml_str = """
+                  profile_models: ensemble_model
+                  """
+
+        args = self._args.copy()
+
+        self.mock_model_config = MockModelConfig(self._model_config_protobuf)
+        self.mock_model_config.start()
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        state_manager = AnalyzerStateManager(config, MagicMock())
+        metrics_manager = MetricsManagerSubclass(
+            config, MagicMock(), MagicMock(), MagicMock(), MagicMock(), state_manager
+        )
+        model_manager = ModelManager(
+            config,
+            MagicMock(),
+            MagicMock(),
+            MagicMock(),
+            metrics_manager,
+            MagicMock(),
+            state_manager,
+            MagicMock(),
+        )
+
+        models = [
+            ConfigModelProfileSpec("ensemble_model"),
+        ]
+        model_manager._check_for_ensemble_model_incompatibility(models)
+
+        self.assertEqual(config.run_config_search_mode, "quick")
+
+    @patch(
+        "model_analyzer.triton.model.model_config.ModelConfig.is_ensemble",
+        return_value=False,
+    )
+    def test_cpu_only_composing_models_error(self, *args):
+        """
+        Test that --cpu-only-composing-models errors when
+        set for non-ensemble/BLS models
+        """
+        yaml_str = """
+                  profile_models: test_model
+                  """
+
+        args = self._args.copy()
+        args.append("--cpu-only-composing-models")
+        args.append("composing_modelA,composing_modelB")
+
+        self.mock_model_config = MockModelConfig(self._model_config_protobuf)
+        self.mock_model_config.start()
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        state_manager = AnalyzerStateManager(config, MagicMock())
+        metrics_manager = MetricsManagerSubclass(
+            config, MagicMock(), MagicMock(), MagicMock(), MagicMock(), state_manager
+        )
+        model_manager = ModelManager(
+            config,
+            MagicMock(),
+            MagicMock(),
+            MagicMock(),
+            metrics_manager,
+            MagicMock(),
+            state_manager,
+            MagicMock(),
+        )
+
+        # RunConfigSearch check
+        models = [
+            ConfigModelProfileSpec("test_model"),
+        ]
+        with self.assertRaises(TritonModelAnalyzerException):
+            model_manager._check_for_ensemble_model_incompatibility(models)
+        self.mock_model_config.stop()
+
+    def _test_model_manager(self, yaml_content, expected_ranges, args=None):
+        """
+        Test helper function that passes the given yaml_str into
+        model_manager, runs the model, and confirms the result is as expected
+        based on a full cartesian product of the lists in the input list of
+        dicts expected_ranges
+        """
+
+        if args is None:
+            args = self._args
+
+        # Use mock model config or else TritonModelAnalyzerException will be thrown as it tries to read from disk
+        self.mock_model_config = MockModelConfig(self._model_config_protobuf)
+        self.mock_model_config.start()
+        config = evaluate_mock_config(args, yaml_content, subcommand="profile")
+
+        state_manager = AnalyzerStateManager(config, MagicMock())
+        metrics_manager = MetricsManagerSubclass(
+            config, MagicMock(), MagicMock(), MagicMock(), MagicMock(), state_manager
+        )
+        model_manager = ModelManager(
+            config,
+            MagicMock(),
+            MagicMock(),
+            MagicMock(),
+            metrics_manager,
+            MagicMock(),
+            state_manager,
+            MagicMock(),
+        )
+
+        model_manager.run_models([config.profile_models[0]])
+        self.mock_model_config.stop()
+
+        self._check_results(model_manager, expected_ranges)
+
+    def _check_results(self, model_manager, expected_ranges):
+        """
+        Create a set of expected and actual run configs and confirm they are equal
+        """
+        run_configs = model_manager._metrics_manager.get_run_configs()
+        expected_configs = MockRunConfigs()
+        expected_configs.populate_from_ranges(expected_ranges)
+
+        self.assertEqual(
+            run_configs.get_configs_set(), expected_configs.get_configs_set()
+        )
+        self.assertEqual(
+            run_configs.get_num_configs(), expected_configs.get_num_configs()
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_variant_name_manager.py b/src/c++/perf_analyzer/tests/test_model_variant_name_manager.py
new file mode 100755
index 000000000..b21e57404
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_model_variant_name_manager.py
@@ -0,0 +1,266 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.config.generate.model_variant_name_manager import (
+    ModelVariantNameManager,
+)
+from model_analyzer.constants import DEFAULT_CONFIG_PARAMS
+from model_analyzer.triton.model.model_config import ModelConfig
+from model_analyzer.triton.model.model_config_variant import ModelConfigVariant
+from tests.common.test_utils import default_encode
+
+from .common import test_result_collector as trc
+
+
+class TestModelVariantNameManager(trc.TestResultCollector):
+    def setUp(self):
+        self._mvnm = ModelVariantNameManager()
+        self._non_default_param_combo = {"foo": 1}
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_default(self):
+        """
+        Check that default config is returned
+        """
+        name = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 1}, DEFAULT_CONFIG_PARAMS
+        )
+
+        self.assertEqual(name, (False, "modelA_config_default"))
+
+    def test_basic(self):
+        """
+        If multiple unique model configs are passed in, the name will keep
+        incrementing
+        """
+        a0 = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 1}, self._non_default_param_combo
+        )
+        a1 = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 2}, self._non_default_param_combo
+        )
+        a2 = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 4}, self._non_default_param_combo
+        )
+
+        self.assertEqual(a0, (False, "modelA_config_0"))
+        self.assertEqual(a1, (False, "modelA_config_1"))
+        self.assertEqual(a2, (False, "modelA_config_2"))
+
+    def test_multiple_models(self):
+        """
+        The two models should have no impact on each other's naming or counts
+        """
+
+        a0 = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 1}, self._non_default_param_combo
+        )
+        b0 = self._mvnm.get_model_variant_name(
+            "modelB", {"A": 1}, self._non_default_param_combo
+        )
+        a1 = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 2}, self._non_default_param_combo
+        )
+        b1 = self._mvnm.get_model_variant_name(
+            "modelB", {"A": 2}, self._non_default_param_combo
+        )
+
+        self.assertEqual(a0, (False, "modelA_config_0"))
+        self.assertEqual(a1, (False, "modelA_config_1"))
+        self.assertEqual(b0, (False, "modelB_config_0"))
+        self.assertEqual(b1, (False, "modelB_config_1"))
+
+    def test_repeat(self):
+        """
+        Calling with the same model name/config/combo multiple times
+        should result in the same config name being returned
+        """
+
+        a0 = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 1}, self._non_default_param_combo
+        )
+        a1 = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 1}, self._non_default_param_combo
+        )
+
+        self.assertEqual(a0, (False, "modelA_config_0"))
+        self.assertEqual(a1, (True, "modelA_config_0"))
+
+    def test_nested_dicts_matching(self):
+        """
+        Test matching with a model config consisting of matching nested dicts
+        """
+        model_config_A_0 = {"A": {"B": {"C": 1, "D": 2}}}
+        model_config_A_1 = {"A": {"B": {"D": 2, "C": 1}}}
+
+        a0 = self._mvnm.get_model_variant_name(
+            "modelA", model_config_A_0, self._non_default_param_combo
+        )
+
+        a1 = self._mvnm.get_model_variant_name(
+            "modelA", model_config_A_1, self._non_default_param_combo
+        )
+
+        self.assertEqual(a0, (False, "modelA_config_0"))
+        self.assertEqual(a1, (True, "modelA_config_0"))
+
+    def test_nested_dicts_different(self):
+        """
+        Test matching with a model config consisting of different nested dicts
+        """
+        model_config_A_0 = {"A": {"B": {"C": 1, "D": 2}}}
+        model_config_A_1 = {"A": {"B": {"D": 1, "C": 2}}}
+
+        a0 = self._mvnm.get_model_variant_name(
+            "modelA", model_config_A_0, self._non_default_param_combo
+        )
+
+        a1 = self._mvnm.get_model_variant_name(
+            "modelA", model_config_A_1, self._non_default_param_combo
+        )
+
+        self.assertEqual(a0, (False, "modelA_config_0"))
+        self.assertEqual(a1, (False, "modelA_config_1"))
+
+    def test_ensemble_default(self):
+        """
+        Test that a default ensemble config is returned
+        """
+        sub_configA = ModelConfigVariant(ModelConfig({}), "modelA_config_default")
+        sub_configB = ModelConfigVariant(ModelConfig({}), "modelB_config_default")
+
+        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
+            [sub_configA, sub_configB]
+        )
+
+        name = self._mvnm.get_ensemble_model_variant_name(
+            "ensemble_model", ensemble_key
+        )
+
+        self.assertEqual(name, (False, "ensemble_model_config_default"))
+
+    def test_ensemble_basic(self):
+        """
+        Test that we can increment the ensemble config numbers
+        """
+        sub_configA = ModelConfigVariant(ModelConfig({}), "modelA_config_0")
+        sub_configB = ModelConfigVariant(ModelConfig({}), "modelB_config_0")
+
+        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
+            [sub_configA, sub_configB]
+        )
+
+        name = self._mvnm.get_ensemble_model_variant_name(
+            "ensemble_model", ensemble_key
+        )
+
+        self.assertEqual(name, (False, "ensemble_model_config_0"))
+
+        sub_configB = ModelConfigVariant(ModelConfig({}), "modelB_config_1")
+
+        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
+            [sub_configA, sub_configB]
+        )
+
+        name = self._mvnm.get_ensemble_model_variant_name(
+            "ensemble_model", ensemble_key
+        )
+
+        self.assertEqual(name, (False, "ensemble_model_config_1"))
+
+        sub_configA = ModelConfigVariant(ModelConfig({}), "modelA_config_1")
+
+        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
+            [sub_configA, sub_configB]
+        )
+
+        name = self._mvnm.get_ensemble_model_variant_name(
+            "ensemble_model", ensemble_key
+        )
+
+        self.assertEqual(name, (False, "ensemble_model_config_2"))
+
+    def test_ensemble_repeat(self):
+        """
+        Calling with the same model name/ensemble key multiple times
+        should result in the same config name being returned
+        """
+        sub_configA = ModelConfigVariant(ModelConfig({}), "modelA_config_0")
+        sub_configB = ModelConfigVariant(ModelConfig({}), "modelB_config_0")
+
+        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
+            [sub_configA, sub_configB]
+        )
+
+        name = self._mvnm.get_ensemble_model_variant_name(
+            "ensemble_model", ensemble_key
+        )
+
+        self.assertEqual(name, (False, "ensemble_model_config_0"))
+
+        name = self._mvnm.get_ensemble_model_variant_name(
+            "ensemble_model", ensemble_key
+        )
+
+        self.assertEqual(name, (True, "ensemble_model_config_0"))
+
+    def test_from_dict(self):
+        """
+        Restoring from a dict should see existing configs
+        """
+        _ = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 1}, self._non_default_param_combo
+        )
+        _ = self._mvnm.get_model_variant_name(
+            "modelB", {"A": 1}, self._non_default_param_combo
+        )
+        _ = self._mvnm.get_model_variant_name(
+            "modelA", {"A": 2}, self._non_default_param_combo
+        )
+
+        mvnm_dict = default_encode(self._mvnm)
+
+        mvnm = ModelVariantNameManager.from_dict(mvnm_dict)
+
+        self.assertEqual(mvnm._model_config_dicts, self._mvnm._model_config_dicts)
+        self.assertEqual(mvnm._model_name_index, self._mvnm._model_name_index)
+
+        a0 = mvnm.get_model_variant_name(
+            "modelA", {"A": 1}, self._non_default_param_combo
+        )
+        b0 = mvnm.get_model_variant_name(
+            "modelB", {"A": 1}, self._non_default_param_combo
+        )
+        a1 = mvnm.get_model_variant_name(
+            "modelA", {"A": 2}, self._non_default_param_combo
+        )
+        b1 = mvnm.get_model_variant_name(
+            "modelB", {"A": 2}, self._non_default_param_combo
+        )
+
+        self.assertEqual(a0, (True, "modelA_config_0"))
+        self.assertEqual(a1, (True, "modelA_config_1"))
+        self.assertEqual(b0, (True, "modelB_config_0"))
+        self.assertEqual(b1, (False, "modelB_config_1"))
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_neighborhood.py b/src/c++/perf_analyzer/tests/test_neighborhood.py
new file mode 100755
index 000000000..ad96899fa
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_neighborhood.py
@@ -0,0 +1,916 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import List
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.generate.coordinate import Coordinate
+from model_analyzer.config.generate.coordinate_data import CoordinateData
+from model_analyzer.config.generate.neighborhood import Neighborhood
+from model_analyzer.config.generate.search_config import NeighborhoodConfig
+from model_analyzer.config.generate.search_dimension import SearchDimension
+from model_analyzer.config.generate.search_dimensions import SearchDimensions
+
+from .common import test_result_collector as trc
+from .common.test_utils import (
+    construct_constraint_manager,
+    construct_run_config_measurement,
+)
+
+
+class TestNeighborhood(trc.TestResultCollector):
+    def setUp(self):
+        self.constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA
+            """
+        )
+
+    # Coordinates can't be sorted by default, but if we convert each one
+    # to a list those can be sorted
+    def _sort_coordinates(self, coordinates: List[Coordinate]) -> List[Coordinate]:
+        sorted_coordinates = coordinates
+        sorted_coordinates.sort(key=lambda c: [x for x in c])
+        return sorted_coordinates
+
+    def tearDown(self):
+        patch.stopall()
+
+    def _construct_rcm(self, throughput: float, latency: float):
+        model_config_name = ["modelA_config_0"]
+
+        # yapf: disable
+        non_gpu_metric_values = [{
+            "perf_throughput": throughput,
+            "perf_latency_p99": latency
+        }]
+        # yapf: enable
+
+        metric_objectives = [{"perf_throughput": 1}]
+        weights = [1]
+
+        rcm = construct_run_config_measurement(
+            model_name="modelA",
+            model_config_names=model_config_name,
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values={},
+            non_gpu_metric_values=non_gpu_metric_values,
+            constraint_manager=self.constraint_manager,
+            metric_objectives=metric_objectives,
+            model_config_weights=weights,
+        )
+        return rcm
+
+    def test_calc_distance(self):
+        a = Coordinate([1, 4, 6, 3])
+        b = Coordinate([4, 2, 6, 0])
+
+        # Euclidean distance is the square root of the
+        # sum of the distances of the coordinates
+        #
+        # Distance = sqrt( (1-4)^2 + (4-2)^2 + (6-6)^2 + (3-0)^2)
+        # Distance = sqrt( 9 + 4 + 0 + 9 )
+        # Distance = sqrt(22)
+        # Distance = 4.69
+        self.assertAlmostEqual(Neighborhood.calc_distance(a, b), 4.69, places=3)
+
+    def test_create_neighborhood(self):
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        n = Neighborhood(
+            nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=CoordinateData()
+        )
+
+        # These are all values within radius of 2 from [1,1,1]
+        # but within the bounds (no negative values)
+        #
+        expected_neighborhood = [
+            [0, 0, 0],
+            [0, 0, 1],
+            [0, 0, 2],
+            [0, 1, 0],
+            [0, 1, 1],
+            [0, 1, 2],
+            [0, 2, 0],
+            [0, 2, 1],
+            [0, 2, 2],
+            [1, 0, 0],
+            [1, 0, 1],
+            [1, 0, 2],
+            [1, 1, 0],
+            [1, 1, 1],
+            [1, 1, 2],
+            [1, 1, 3],
+            [1, 2, 0],
+            [1, 2, 1],
+            [1, 2, 2],
+            [1, 3, 1],
+            [2, 0, 0],
+            [2, 0, 1],
+            [2, 0, 2],
+            [2, 1, 0],
+            [2, 1, 1],
+            [2, 1, 2],
+            [2, 2, 0],
+            [2, 2, 1],
+            [2, 2, 2],
+            [3, 1, 1],
+        ]
+
+        expected_coordinates = [Coordinate(x) for x in expected_neighborhood]
+
+        self.assertEqual(
+            self._sort_coordinates(n._neighborhood),
+            self._sort_coordinates(expected_coordinates),
+        )
+
+    def test_large_neighborhood(self):
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("a1", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("b1", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("c1", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+        dims.add_dimensions(
+            1,
+            [
+                SearchDimension("a2", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("b2", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("c2", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+        nc = NeighborhoodConfig(dims, radius=3, min_initialized=3)
+        n = Neighborhood(
+            nc,
+            home_coordinate=Coordinate([1, 1, 1, 1, 1, 1]),
+            coordinate_data=CoordinateData(),
+        )
+
+        self.assertEqual(2328, len(n._neighborhood))
+
+    def test_num_initialized(self):
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
+
+        rcm = self._construct_rcm(throughput=100, latency=80)
+
+        # Start with 0 initialized
+        self.assertEqual(0, len(n._get_coordinates_with_valid_measurements()))
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Home coordinate is ignored. No change to num initialized/visited.
+        cd.set_measurement(Coordinate([1, 1, 1]), rcm)
+        self.assertEqual(0, len(n._get_coordinates_with_valid_measurements()))
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Incremented to 1 initialized
+        cd.set_measurement(Coordinate([0, 0, 0]), rcm)
+        self.assertEqual(1, len(n._get_coordinates_with_valid_measurements()))
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Set same point. No change to num initialized
+        cd.set_measurement(Coordinate([0, 0, 0]), rcm)
+        self.assertEqual(1, len(n._get_coordinates_with_valid_measurements()))
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Set a point outside of neighborhood
+        cd.set_measurement(Coordinate([0, 0, 4]), rcm)
+        self.assertEqual(1, len(n._get_coordinates_with_valid_measurements()))
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Set a third point inside of neighborhood
+        cd.set_measurement(Coordinate([1, 0, 0]), rcm)
+        self.assertEqual(2, len(n._get_coordinates_with_valid_measurements()))
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Set the none measurement. No changed to num initialized.
+        cd.set_measurement(Coordinate([0, 1, 1]), None)
+        self.assertEqual(2, len(n._get_coordinates_with_valid_measurements()))
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Set the last point inside of neighborhood
+        cd.set_measurement(Coordinate([1, 1, 0]), rcm)
+        self.assertEqual(3, len(n._get_coordinates_with_valid_measurements()))
+        self.assertTrue(n.enough_coordinates_initialized())
+
+    def test_get_all_adjacent_neighbors(self):
+        """
+        Test that _get_all_adjacent_neighbors() works, and understands dimension bounds
+
+        For this test, home is [0,1,4].
+
+        The possible adjacent neighbors are:
+          [-1,1,4], [1,1,4], [0,0,4], [0,2,4], [0,1,3], [0,1,5]
+
+        However, [-1,1,4] and [0,1,5] are outside of the dimension bounds and should not
+        be part of the returned list
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("a", SearchDimension.DIMENSION_TYPE_LINEAR, min=0),
+                SearchDimension("b", SearchDimension.DIMENSION_TYPE_LINEAR, min=0),
+                SearchDimension(
+                    "c", SearchDimension.DIMENSION_TYPE_LINEAR, min=1, max=4
+                ),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        n = Neighborhood(
+            nc, home_coordinate=Coordinate([0, 1, 4]), coordinate_data=CoordinateData()
+        )
+
+        adjacent_neighbors = n._get_all_adjacent_neighbors()
+        expected_list = [
+            Coordinate([1, 1, 4]),
+            Coordinate([0, 0, 4]),
+            Coordinate([0, 2, 4]),
+            Coordinate([0, 1, 3]),
+        ]
+
+        self.assertEqual(adjacent_neighbors, expected_list)
+
+    def test_num_initialized_slow_mode(self):
+        """
+        Test that _enough_coordinates_initialized() works in slow mode
+
+        Start with home=[0,0,1] in slow mode.
+        Only once all of the adjacent neighbors are added should
+        enough_coordinates_initialized() return true
+
+        Those adjacent neighbors are [1,0,1], [0,1,1], [0,0,0], [0,0,2]
+        """
+
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0, 1]), coordinate_data=cd)
+        n.force_slow_mode()
+
+        # Start with None initialized
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Home coordinate is ignored
+        cd.set_measurement(Coordinate([0, 0, 1]), MagicMock())
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Add 3 of the 4 neighbors
+        cd.set_measurement(Coordinate([1, 0, 1]), MagicMock())
+        self.assertFalse(n.enough_coordinates_initialized())
+        cd.set_measurement(Coordinate([0, 1, 1]), MagicMock())
+        self.assertFalse(n.enough_coordinates_initialized())
+        cd.set_measurement(Coordinate([0, 0, 0]), MagicMock())
+        self.assertFalse(n.enough_coordinates_initialized())
+
+        # Add the final neighbor
+        cd.set_measurement(Coordinate([0, 0, 2]), MagicMock())
+        self.assertTrue(n.enough_coordinates_initialized())
+
+    def test_is_slow_mode(self):
+        """
+        Test that _is_slow_mode() only returns true by these 3 conditions:
+        - force_slow_mode() has been called
+        - Home is passing and anyone in the neighborhood is failing
+        - Home is failing and anyone in the neighborhood is passing
+
+        Test this by running a number of sub-tests with different configurations
+        of num_passing, num_failing, home_passing, and home_measured
+        """
+
+        # Home not measured -> False
+        self._test_is_slow_mode_helper(home_measured=False, expected_result=False)
+
+        # Home passing, all passing -> False
+        self._test_is_slow_mode_helper(
+            home_passing=True, num_passing=3, num_failing=0, expected_result=False
+        )
+
+        # Same as previous, but force_slow_mode is called -> True
+        self._test_is_slow_mode_helper(
+            force_slow=True,
+            home_passing=True,
+            num_passing=3,
+            num_failing=0,
+            expected_result=True,
+        )
+
+        # Home passing, some failing -> True
+        self._test_is_slow_mode_helper(
+            num_passing=3, num_failing=1, home_passing=True, expected_result=True
+        )
+
+        # Home passing, all failing -> True
+        self._test_is_slow_mode_helper(
+            num_passing=0, num_failing=3, home_passing=True, expected_result=True
+        )
+
+        # Home failing, all failing -> False
+        self._test_is_slow_mode_helper(
+            num_passing=0, num_failing=3, home_passing=False, expected_result=False
+        )
+
+        # Home failing, some passing -> True
+        self._test_is_slow_mode_helper(
+            num_passing=1, num_failing=3, home_passing=False, expected_result=True
+        )
+
+        # Home failing, all passing -> True
+        self._test_is_slow_mode_helper(
+            num_passing=3, num_failing=0, home_passing=False, expected_result=True
+        )
+
+    def _test_is_slow_mode_helper(
+        self,
+        home_measured=True,
+        num_passing=1,
+        num_failing=1,
+        home_passing=True,
+        force_slow=False,
+        expected_result=False,
+    ):
+        total_measurements = num_failing + num_passing
+
+        passing_ret_val = [[0] * num_passing, []]
+        all_ret_val = [[0] * total_measurements, []]
+
+        patches = []
+        patches.append(
+            patch.object(
+                Neighborhood, "_is_home_measured", MagicMock(return_value=home_measured)
+            )
+        )
+        patches.append(
+            patch.object(
+                Neighborhood,
+                "_get_measurements_passing_constraints",
+                MagicMock(return_value=passing_ret_val),
+            )
+        )
+        patches.append(
+            patch.object(
+                Neighborhood,
+                "_get_all_measurements",
+                MagicMock(return_value=all_ret_val),
+            )
+        )
+        patches.append(
+            patch.object(
+                Neighborhood,
+                "_is_home_passing_constraints",
+                MagicMock(return_value=home_passing),
+            )
+        )
+
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        n = Neighborhood(
+            nc, home_coordinate=Coordinate([1, 1]), coordinate_data=CoordinateData()
+        )
+
+        for p in patches:
+            p.start()
+
+        if force_slow:
+            n.force_slow_mode()
+
+        self.assertEqual(n._is_slow_mode(), expected_result)
+
+        for p in patches:
+            p.stop()
+
+    def test_get_all_measurements(self):
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
+
+        rcm0 = self._construct_rcm(throughput=100, latency=50)
+        rcm1 = self._construct_rcm(throughput=700, latency=350)
+        rcm2 = self._construct_rcm(throughput=300, latency=130)
+
+        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
+        cd.set_measurement(Coordinate([2, 1, 1]), rcm1)
+        cd.set_measurement(Coordinate([1, 2, 1]), rcm2)
+        cd.set_measurement(Coordinate([1, 2, 2]), None)
+
+        expected_vectors = [
+            Coordinate([1, 2, 1]) - Coordinate([1, 1, 1]),
+            Coordinate([2, 1, 1]) - Coordinate([1, 1, 1]),
+        ]
+        expected_measurements = [rcm2, rcm1]
+
+        vectors, measurements = n._get_all_measurements()
+        for ev, em in zip(expected_vectors, expected_measurements):
+            self.assertIn(ev, vectors)
+            self.assertIn(em, measurements)
+
+    def test_get_constraints_passing_measurements(self):
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
+
+        # Constraints:
+        #   - Minimum throughput of 100 infer/sec
+        #   - Maximum latency of 300 ms
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 100
+                  perf_latency_p99:
+                    max: 300
+            """
+        )
+
+        rcm0 = self._construct_rcm(throughput=100, latency=50)  # pass
+        rcm0.set_constraint_manager(constraint_manager)
+
+        rcm1 = self._construct_rcm(throughput=700, latency=350)  # fail
+        rcm1.set_constraint_manager(constraint_manager)
+
+        rcm2 = self._construct_rcm(throughput=300, latency=130)  # pass
+        rcm2.set_constraint_manager(constraint_manager)
+
+        rcm3 = self._construct_rcm(throughput=400, latency=290)  # pass
+        rcm3.set_constraint_manager(constraint_manager)
+
+        rcm4 = self._construct_rcm(throughput=850, latency=400)  # fail
+        rcm4.set_constraint_manager(constraint_manager)
+
+        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
+        cd.set_measurement(Coordinate([2, 1, 1]), rcm1)
+        cd.set_measurement(Coordinate([1, 2, 1]), rcm2)
+        cd.set_measurement(Coordinate([1, 1, 2]), rcm3)
+        cd.set_measurement(Coordinate([1, 2, 2]), rcm4)
+
+        expected_vectors = [
+            Coordinate([1, 1, 2]) - Coordinate([1, 1, 1]),
+            Coordinate([1, 2, 1]) - Coordinate([1, 1, 1]),
+        ]
+        expected_measurements = [rcm3, rcm2]
+
+        vectors, measurements = n._get_measurements_passing_constraints()
+        for ev, em in zip(expected_vectors, expected_measurements):
+            self.assertIn(ev, vectors)
+            self.assertIn(em, measurements)
+
+    def test_step_vector_both_home_and_neighbors_passing_constraints(self):
+        """
+        Test _get_step_vector method where home and neighbors are passing
+
+          1. Get all vectors and measurements
+                [2, 3, 1] - [1, 1, 1] = [1, 2, 0] (throughput=300, latency=130)
+                [1, 1, 2] - [1, 1, 1] = [0, 0, 1] (throughput=400, latency=290)
+
+          2. Apply the objective-comparison weights to any non-zero dimensions
+                [1,2,0] with weight 1.0 -> [1.0, 1.0, 0.0]
+                [0,0,1] with weight 1.2 -> [0.0, 0.0, 1.2]
+
+          3. Sum the results
+                [1.0, 1.0, 0.0] + [0.0, 0.0, 1.2] = [1.0, 1.0, 1.2]
+
+          4. Divide by the sum of the vectors:
+                [1.0, 1.0, 1.2] / [1,2,1] = [1.0, 0.5, 1.2]
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=3, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
+
+        # Constraints:
+        #   - Minimum throughput of 100 infer/sec
+        #   - Maximum latency of 300 ms
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 100
+                  perf_latency_p99:
+                    max: 300
+            """
+        )
+
+        rcm0 = self._construct_rcm(throughput=100, latency=50)  # pass
+        rcm0.set_constraint_manager(constraint_manager)
+
+        rcm1 = self._construct_rcm(throughput=300, latency=130)  # pass
+        rcm1.set_constraint_manager(constraint_manager)
+
+        rcm2 = self._construct_rcm(throughput=400, latency=290)  # pass
+        rcm2.set_constraint_manager(constraint_manager)
+
+        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
+        cd.set_measurement(Coordinate([2, 3, 1]), rcm1)
+        cd.set_measurement(Coordinate([1, 1, 2]), rcm2)
+
+        hm = cd.get_measurement(Coordinate([1, 1, 1]))
+        self.assertTrue(hm.is_passing_constraints())
+
+        step_vector = n._get_step_vector()
+        expected_step_vector = Coordinate([1.0, 0.5, 1.2])
+
+        self.assertEqual(step_vector, expected_step_vector)
+
+    def test_step_vector_both_home_and_neighbors_failing_constraints(self):
+        """
+        Test _get_step_vector method where home and neighbors are failing
+
+          1. Get all vectors and measurements
+                Home latency is 450
+                [3, 2, 1] - [1, 1, 1] = [2, 1, 0] (latency=360)
+                [1, 2, 0] - [1, 1, 1] = [0, 1, -1] (latency=540)
+
+          2. Apply the constraint-comparison weights to any non-zero dimensions,
+             inverting the weight if coordinate is negative
+                [2,1,0] with weight 0.3 -> [0.3, 0.3, 0.0]
+                [0,1,-1] with weight -0.3 -> [0.0, -0.3, 0.3]
+
+          3. Sum the results
+                [0.3, 0.3, 0.0] + [0.0, -0.3, 0.3] = [0.3, 0.0, 0.3]
+
+          4. Divide by the sum of the absolute value of the vectors:
+                [0.3, 0.0, 0.3] / [2,2,1] = [0.15, 0.0, 0.3]
+
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=3, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
+
+        # Constraints:
+        #   - Minimum throughput of 100 infer/sec
+        #   - Maximum latency of 300 ms
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 100
+                  perf_latency_p99:
+                    max: 300
+            """
+        )
+
+        rcm0 = self._construct_rcm(throughput=500, latency=450)  # fail
+        rcm0.set_constraint_manager(constraint_manager)
+
+        rcm1 = self._construct_rcm(throughput=700, latency=360)  # fail
+        rcm1.set_constraint_manager(constraint_manager)
+
+        rcm2 = self._construct_rcm(throughput=850, latency=540)  # fail
+        rcm2.set_constraint_manager(constraint_manager)
+
+        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
+        cd.set_measurement(Coordinate([3, 2, 1]), rcm1)
+        cd.set_measurement(Coordinate([1, 2, 0]), rcm2)
+
+        hm = cd.get_measurement(Coordinate([1, 1, 1]))
+        self.assertFalse(hm.is_passing_constraints())
+
+        step_vector = n._get_step_vector()
+        expected_step_vector = Coordinate([0.15, 0.0, 0.3])
+        self.assertEqual(step_vector, expected_step_vector)
+
+    def test_determine_new_home_one_dimension(self):
+        """
+        Test determine_new_home for a case where only
+        one of the dimensions increases the measurement.
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0]), coordinate_data=cd)
+
+        rcm0 = self._construct_rcm(throughput=1, latency=5)
+        rcm1 = self._construct_rcm(throughput=3, latency=5)
+        rcm2 = self._construct_rcm(throughput=1, latency=5)
+
+        cd.set_measurement(Coordinate([0, 0]), rcm0)  # home coordinate
+        cd.set_measurement(Coordinate([1, 0]), rcm1)
+        cd.set_measurement(Coordinate([0, 1]), rcm2)
+
+        new_coord = n.determine_new_home()
+        self.assertEqual(new_coord, Coordinate([2, 0]))
+
+    def test_determine_new_home_two_dimensions(self):
+        """
+        Test determine_new_home for a case where both of the
+        dimensions increases the measurement equally.
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0]), coordinate_data=cd)
+
+        rcm0 = self._construct_rcm(throughput=1, latency=5)
+        rcm1 = self._construct_rcm(throughput=3, latency=5)
+        rcm2 = self._construct_rcm(throughput=3, latency=5)
+
+        cd.set_measurement(Coordinate([0, 0]), rcm0)  # home coordinate
+        cd.set_measurement(Coordinate([1, 0]), rcm1)
+        cd.set_measurement(Coordinate([0, 1]), rcm2)
+
+        new_coord = n.determine_new_home()
+        self.assertEqual(new_coord, Coordinate([2, 2]))
+
+    def test_determine_new_home(self):
+        """
+        Test determine_new_home for a case where both of the
+        dimensions increases the measurement equally
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0]), coordinate_data=cd)
+
+        rcm0 = self._construct_rcm(throughput=1, latency=5)
+        rcm1 = self._construct_rcm(throughput=7, latency=5)
+        rcm2 = self._construct_rcm(throughput=7, latency=5)
+
+        cd.set_measurement(Coordinate([0, 0]), rcm0)  # home coordinate
+        cd.set_measurement(Coordinate([1, 0]), rcm1)
+        cd.set_measurement(Coordinate([0, 1]), rcm2)
+
+        # 7x increase in throughputs will result in the maximum step of [3,3]
+        new_coord = n.determine_new_home()
+        self.assertEqual(new_coord, Coordinate([3, 3]))
+
+        new_coord = n.determine_new_home()
+        self.assertEqual(new_coord, Coordinate([3, 3]))
+
+    def test_determine_new_home_out_of_bounds(self):
+        """
+        Test that determine_new_home will clamp the result to
+        the search dimension bounds.
+
+        Both dimensions are defined to only be from 2-7. The test sets up
+        the case where the next step WOULD be from [3,6] -> [0,9] if not
+        for bounding into the defined range.
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension(
+                    "foo", SearchDimension.DIMENSION_TYPE_LINEAR, min=2, max=7
+                ),
+                SearchDimension(
+                    "bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL, min=2, max=7
+                ),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=8, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([3, 6]), coordinate_data=cd)
+
+        rcm0 = self._construct_rcm(throughput=1, latency=5)
+        rcm1 = self._construct_rcm(throughput=7, latency=5)
+
+        cd.set_measurement(Coordinate([3, 6]), rcm0)  # home coordinate
+        cd.set_measurement(Coordinate([2, 7]), rcm1)
+
+        new_coord = n.determine_new_home()
+        self.assertEqual(new_coord, Coordinate([2, 7]))
+
+    def test_all_same_throughputs(self):
+        """
+        Test that when all the coordinates in the neighborhood has the
+        same throughputs, the step vector is zero and new coordinate is
+        same as the home coordinate.
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        nc = NeighborhoodConfig(dims, radius=3, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
+
+        rcm0 = self._construct_rcm(throughput=10, latency=5)
+        rcm1 = self._construct_rcm(throughput=10, latency=5)
+        rcm2 = self._construct_rcm(throughput=10, latency=5)
+        rcm3 = self._construct_rcm(throughput=10, latency=5)
+
+        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
+        cd.set_measurement(Coordinate([1, 0, 0]), rcm1)
+        cd.set_measurement(Coordinate([0, 1, 0]), rcm2)
+        cd.set_measurement(Coordinate([0, 0, 1]), rcm3)
+
+        step_vector = n._get_step_vector()
+        self.assertEqual(step_vector, Coordinate([0, 0, 0]))
+
+        n.force_slow_mode()
+        new_coord = n.determine_new_home()
+        self.assertEqual(new_coord, Coordinate([1, 1, 1]))
+
+    def test_translate_step_vector(self):
+        """
+        Test the functionality of translate_step_vector()
+
+        Any dimension value that is less than translation_list[0] will get 0
+
+        Any dimension value that is between translation_list[i] and translation_list[i+1]
+        will get i+1
+        """
+
+        # Test generic
+        self._test_translate_step_vector_helper(
+            input_vector=[0.35, 0.25, 0.15, 0.05],
+            translation_list=[0.1, 0.2, 0.3],
+            expected_vector=[3, 2, 1, 0],
+        )
+
+        # Test negative
+        self._test_translate_step_vector_helper(
+            input_vector=[-0.35, -0.25, -0.15, -0.05],
+            translation_list=[0.1, 0.2, 0.3],
+            expected_vector=[-3, -2, -1, 0],
+        )
+
+        # Test boundary case. Exact match rounds down
+        self._test_translate_step_vector_helper(
+            input_vector=[0.099, 0.100, 0.101],
+            translation_list=[0.1],
+            expected_vector=[0, 0, 1],
+        )
+
+    def _test_translate_step_vector_helper(
+        self, input_vector, translation_list, expected_vector
+    ):
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0, [SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR)]
+        )
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([0]), coordinate_data=cd)
+
+        input_coord = Coordinate(input_vector)
+        expected_coord = Coordinate(expected_vector)
+        output_coord = n._translate_step_vector(input_coord, translation_list)
+
+        self.assertEqual(expected_coord, output_coord)
+
+    def _test_calculate_step_vector_from_vectors_and_weights(self):
+        """
+        Test the functionality of __test_calculate_step_vector_from_vectors_and_weights()
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
+        cd = CoordinateData()
+        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0, 0]), coordinate_data=cd)
+
+        vectors = [[0, 0, 1], [1, -3, 2], [-4, 2, 0], [3, 2, 6]]
+        weights = [0.5, -3.0, -1, 0]
+
+        # Sum up weights into vectors:
+        # [0,0,0.5] + [-3,3,-3] + [1, -1, 0] + [0,0,0] = [-2, 2, -2.5]
+        #
+        # Sum up absolute value of vectors:
+        # [0,0,1] + [1,3,2] + [4,2,0] + [3,2,6] = [8, 7, 9]
+        #
+        # Divide:
+        # [-2, 2, -2.5] / [8, 7, 9] = [-2/8, 2/7, -2.5/9]
+
+        step_vector = n._calculate_step_vector_from_vectors_and_weights(
+            vectors, weights
+        )
+        expected_result = [-2 / 8, 2 / 7, -2.5 / 9]
+        self.assertEqual(step_vector, expected_result)
diff --git a/src/c++/perf_analyzer/tests/test_parameter_search.py b/src/c++/perf_analyzer/tests/test_parameter_search.py
new file mode 100755
index 000000000..7f410bb26
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_parameter_search.py
@@ -0,0 +1,336 @@
+#!/usr/bin/env python3
+
+# Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from math import log2
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.input.config_command_profile import ConfigCommandProfile
+from model_analyzer.config.input.config_defaults import (
+    DEFAULT_RUN_CONFIG_MAX_CONCURRENCY,
+    DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE,
+    DEFAULT_RUN_CONFIG_MIN_CONCURRENCY,
+    DEFAULT_RUN_CONFIG_MIN_REQUEST_RATE,
+)
+from model_analyzer.constants import THROUGHPUT_MINIMUM_CONSECUTIVE_PARAMETER_TRIES
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.result.constraint_manager import ConstraintManager
+from model_analyzer.result.parameter_search import ParameterSearch
+from model_analyzer.result.run_config_measurement import RunConfigMeasurement
+
+from .common import test_result_collector as trc
+from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
+
+
+class TestParameterSearch(trc.TestResultCollector):
+    def setUp(self):
+        self._min_concurrency_index = int(log2(DEFAULT_RUN_CONFIG_MIN_CONCURRENCY))
+        self._max_concurrency_index = int(log2(DEFAULT_RUN_CONFIG_MAX_CONCURRENCY))
+
+        self._expected_concurrencies = [
+            2**c
+            for c in range(self._min_concurrency_index, self._max_concurrency_index + 1)
+        ]
+        self._concurrencies = []
+
+        self._min_request_rate_index = int(log2(DEFAULT_RUN_CONFIG_MIN_REQUEST_RATE))
+        self._max_request_rate_index = int(log2(DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE))
+
+        self._expected_request_rates = [
+            2**rr
+            for rr in range(
+                self._min_request_rate_index, self._max_request_rate_index + 1
+            )
+        ]
+        self._request_rates = []
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_concurrency_sweep(self):
+        """
+        Test sweeping concurrency from min to max, when no constraints are present
+        and throughput is linearly increasing
+        """
+        config = self._create_single_model_no_constraints()
+        constraint_manager = ConstraintManager(config)
+        concurrency_search = ParameterSearch(config)
+
+        for concurrency in concurrency_search.search_parameters():
+            self._concurrencies.append(concurrency)
+
+            concurrency_search.add_run_config_measurement(
+                run_config_measurement=self._construct_rcm(
+                    throughput=100 * concurrency,
+                    latency=10,
+                    concurrency=concurrency,
+                    constraint_manager=constraint_manager,
+                )
+            )
+
+        self.assertEqual(self._concurrencies, self._expected_concurrencies)
+
+    def test_request_rate_sweep(self):
+        """
+        Test sweeping request rate from min to max, when no constraints are present
+        and throughput is linearly increasing
+        """
+        config = self._create_single_model_no_constraints()
+        constraint_manager = ConstraintManager(config)
+        concurrency_search = ParameterSearch(
+            config, model_parameters={"request_rate": "True"}
+        )
+
+        for request_rate in concurrency_search.search_parameters():
+            self._request_rates.append(request_rate)
+
+            concurrency_search.add_run_config_measurement(
+                run_config_measurement=self._construct_rcm(
+                    throughput=100 * request_rate,
+                    latency=10,
+                    request_rate=request_rate,
+                    constraint_manager=constraint_manager,
+                )
+            )
+
+        self.assertEqual(self._request_rates, self._expected_request_rates)
+
+    def test_saturating_sweep(self):
+        """
+        Test sweeping concurrency from min to max, when no constraints are present
+        and throughput increases and then saturates
+        """
+        config = self._create_single_model_no_constraints()
+        constraint_manager = ConstraintManager(config)
+        concurrency_search = ParameterSearch(config)
+        INCREASE_THROUGHPUT_COUNT = 4
+
+        # [100, 200, 400, 800, 1000, 1000,...]
+        throughputs = [
+            100 * 2**c if c < INCREASE_THROUGHPUT_COUNT else 1000
+            for c in range(self._min_concurrency_index, self._max_concurrency_index + 1)
+        ]
+
+        for i, concurrency in enumerate(concurrency_search.search_parameters()):
+            self._concurrencies.append(concurrency)
+
+            concurrency_search.add_run_config_measurement(
+                run_config_measurement=self._construct_rcm(
+                    throughput=throughputs[i],
+                    latency=10,
+                    concurrency=concurrency,
+                    constraint_manager=constraint_manager,
+                )
+            )
+
+        expected_concurrencies = [
+            2**c
+            for c in range(
+                INCREASE_THROUGHPUT_COUNT
+                + THROUGHPUT_MINIMUM_CONSECUTIVE_PARAMETER_TRIES
+            )
+        ]
+        self.assertEqual(self._concurrencies, expected_concurrencies)
+
+    def test_sweep_with_constraints_decreasing(self):
+        """
+        Test sweeping concurrency from min to max, with 95ms latency constraint
+        and throughput is linearly increasing - which causes a decreasing binary search
+        """
+        config = self._create_single_model_with_constraints("95")
+        constraint_manager = ConstraintManager(config)
+        concurrency_search = ParameterSearch(config)
+
+        self._expected_concurrencies.extend([12, 10, 9])
+        latencies = [10 * c for c in self._expected_concurrencies]
+
+        for i, concurrency in enumerate(concurrency_search.search_parameters()):
+            self._concurrencies.append(concurrency)
+
+            concurrency_search.add_run_config_measurement(
+                run_config_measurement=self._construct_rcm(
+                    throughput=100 * concurrency,
+                    latency=latencies[i],
+                    concurrency=concurrency,
+                    constraint_manager=constraint_manager,
+                )
+            )
+
+        self.assertEqual(self._concurrencies, self._expected_concurrencies)
+
+    def test_sweep_with_constraints_decrease_then_increase(self):
+        """
+        Test sweeping concurrency from min to max, with 155ms latency constraint
+        and throughput is linearly increasing - which causes a decreasing, then increasing binary search
+        """
+        config = self._create_single_model_with_constraints("155")
+        constraint_manager = ConstraintManager(config)
+        concurrency_search = ParameterSearch(config)
+
+        self._expected_concurrencies.extend([12, 14, 15])
+        latencies = [10 * c for c in self._expected_concurrencies]
+
+        for i, concurrency in enumerate(concurrency_search.search_parameters()):
+            self._concurrencies.append(concurrency)
+
+            concurrency_search.add_run_config_measurement(
+                run_config_measurement=self._construct_rcm(
+                    throughput=100 * concurrency,
+                    latency=latencies[i],
+                    concurrency=concurrency,
+                    constraint_manager=constraint_manager,
+                )
+            )
+
+        self.assertEqual(self._concurrencies, self._expected_concurrencies)
+
+    def test_sweep_with_multiple_violation_areas(self):
+        """
+        Test sweeping concurrency from min to max, with 155ms latency constraint
+        with violations in two separate locations
+        """
+        config = self._create_single_model_with_constraints("155")
+        constraint_manager = ConstraintManager(config)
+        concurrency_search = ParameterSearch(config)
+
+        self._expected_concurrencies.extend([12, 14, 15])
+        latencies = [10 * c for c in self._expected_concurrencies]
+        # this adds an early constraint violation which should be ignored
+        latencies[1] = 200
+
+        for i, concurrency in enumerate(concurrency_search.search_parameters()):
+            self._concurrencies.append(concurrency)
+
+            concurrency_search.add_run_config_measurement(
+                run_config_measurement=self._construct_rcm(
+                    throughput=100 * concurrency,
+                    latency=latencies[i],
+                    concurrency=concurrency,
+                    constraint_manager=constraint_manager,
+                )
+            )
+
+        self.assertEqual(self._concurrencies, self._expected_concurrencies)
+
+    def test_sweep_with_constraints_hitting_limit(self):
+        """
+        Test sweeping concurrency from min to max, with 970ms latency constraint
+        and throughput matches concurrency, this will cause BCS to
+        quit after DEFAULT_RUN_CONFIG_MAX_BINARY_SEARCH_STEPS (5) attempts
+        """
+        config = self._create_single_model_with_constraints("970")
+        constraint_manager = ConstraintManager(config)
+        concurrency_search = ParameterSearch(config)
+
+        self._expected_concurrencies.extend([768, 896, 960, 992, 976])
+        latencies = self._expected_concurrencies
+
+        for i, concurrency in enumerate(concurrency_search.search_parameters()):
+            self._concurrencies.append(concurrency)
+
+            concurrency_search.add_run_config_measurement(
+                run_config_measurement=self._construct_rcm(
+                    throughput=100 * concurrency,
+                    latency=latencies[i],
+                    concurrency=concurrency,
+                    constraint_manager=constraint_manager,
+                )
+            )
+
+        self.assertEqual(self._concurrencies, self._expected_concurrencies)
+
+    def test_not_adding_measurements(self):
+        """
+        Test that an exception is raised if measurements are not added
+        """
+        config = self._create_single_model_no_constraints()
+        constraint_manager = ConstraintManager(config)
+        concurrency_search = ParameterSearch(config)
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            for concurrency in concurrency_search.search_parameters():
+                self._concurrencies.append(concurrency)
+
+                if concurrency < 32:
+                    concurrency_search.add_run_config_measurement(
+                        run_config_measurement=self._construct_rcm(
+                            throughput=100 * concurrency,
+                            latency=10,
+                            concurrency=concurrency,
+                            constraint_manager=constraint_manager,
+                        )
+                    )
+
+    def _create_single_model_no_constraints(self):
+        args = ["model-analyzer", "profile", "--profile-models", "test_model"]
+        yaml_str = ""
+        config = evaluate_mock_config(args, yaml_str)
+
+        return config
+
+    def _create_single_model_with_constraints(
+        self, latency_budget: str
+    ) -> ConfigCommandProfile:
+        args = [
+            "model-analyzer",
+            "profile",
+            "--profile-models",
+            "test_model",
+            "--latency-budget",
+            latency_budget,
+        ]
+        yaml_str = ""
+        config = evaluate_mock_config(args, yaml_str)
+
+        return config
+
+    def _construct_rcm(
+        self,
+        throughput: int,
+        latency: int,
+        constraint_manager: ConstraintManager,
+        concurrency: int = 0,
+        request_rate: int = 0,
+    ) -> RunConfigMeasurement:
+        if concurrency:
+            self.model_specific_pa_params = [
+                {"batch_size": 1, "concurrency": concurrency}
+            ]
+        else:
+            self.model_specific_pa_params = [
+                {"batch_size": 1, "request_rate": request_rate}
+            ]
+
+        self.rcm0_non_gpu_metric_values = [
+            {
+                "perf_throughput": throughput,
+                "perf_latency_p99": latency,
+                "cpu_used_ram": 1000,
+            }
+        ]
+
+        return construct_run_config_measurement(
+            model_name="test_model",
+            model_config_names=["test_model_config_0"],
+            model_specific_pa_params=self.model_specific_pa_params,
+            gpu_metric_values=MagicMock(),
+            non_gpu_metric_values=self.rcm0_non_gpu_metric_values,
+            constraint_manager=constraint_manager,
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_pdf_report.py b/src/c++/perf_analyzer/tests/test_pdf_report.py
new file mode 100755
index 000000000..acd938aed
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_pdf_report.py
@@ -0,0 +1,61 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.reports.pdf_report import PDFReport
+
+from .common import test_result_collector as trc
+
+
+class TestPDFReportMethods(trc.TestResultCollector):
+    """
+    Tests the methods of the PDFReport class
+    """
+
+    def setUp(self):
+        self.maxDiff = None
+        self.report = PDFReport()
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_write_report(self):
+        with patch(
+            "model_analyzer.reports.pdf_report.pdfkit", MagicMock()
+        ) as pdfkit_mock:
+            self.report.add_title("Test PDF Report")
+            self.report.add_subheading("Throughput vs. Latency")
+            test_paragraph = (
+                "This is a test paragraph with a lot to say."
+                " There is more than one line in this paragraph."
+            )
+            self.report.add_paragraph(test_paragraph, font_size=14)
+            self.report.write_report("test_report_filename")
+
+            expected_report_body = (
+                "<html><head><style></style></head><body><center><h1>Test PDF Report</h1></center><h3>Throughput vs. Latency</h3>"
+                f'<div style="font-size:14"><p>{test_paragraph}</p></div>'
+                "</body></html>"
+            )
+            pdfkit_mock.from_string.assert_called_with(
+                expected_report_body, "test_report_filename", options={"quiet": ""}
+            )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_perf_analyzer.py b/src/c++/perf_analyzer/tests/test_perf_analyzer.py
new file mode 100755
index 000000000..e95f0d4a1
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_perf_analyzer.py
@@ -0,0 +1,873 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, mock_open, patch
+
+from model_analyzer.config.run.model_run_config import ModelRunConfig
+from model_analyzer.config.run.run_config import RunConfig
+from model_analyzer.constants import (
+    MEASUREMENT_REQUEST_COUNT_STEP,
+    MEASUREMENT_WINDOW_STEP,
+    PERF_ANALYZER_MEASUREMENT_WINDOW,
+    PERF_ANALYZER_MINIMUM_REQUEST_COUNT,
+)
+from model_analyzer.device.gpu_device import GPUDevice
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.perf_analyzer.perf_analyzer import PerfAnalyzer
+from model_analyzer.perf_analyzer.perf_config import PerfAnalyzerConfig
+from model_analyzer.record.types.gpu_free_memory import GPUFreeMemory
+from model_analyzer.record.types.gpu_power_usage import GPUPowerUsage
+from model_analyzer.record.types.gpu_used_memory import GPUUsedMemory
+from model_analyzer.record.types.gpu_utilization import GPUUtilization
+from model_analyzer.record.types.perf_client_response_wait import PerfClientResponseWait
+from model_analyzer.record.types.perf_client_send_recv import PerfClientSendRecv
+from model_analyzer.record.types.perf_latency_avg import PerfLatencyAvg
+from model_analyzer.record.types.perf_latency_p90 import PerfLatencyP90
+from model_analyzer.record.types.perf_latency_p95 import PerfLatencyP95
+from model_analyzer.record.types.perf_latency_p99 import PerfLatencyP99
+from model_analyzer.record.types.perf_server_compute_infer import PerfServerComputeInfer
+from model_analyzer.record.types.perf_server_compute_input import PerfServerComputeInput
+from model_analyzer.record.types.perf_server_compute_output import (
+    PerfServerComputeOutput,
+)
+from model_analyzer.record.types.perf_server_queue import PerfServerQueue
+from model_analyzer.record.types.perf_throughput import PerfThroughput
+from model_analyzer.triton.client.client_factory import TritonClientFactory
+from model_analyzer.triton.server.server_config import TritonServerConfig
+from model_analyzer.triton.server.server_factory import TritonServerFactory
+
+from .common import test_result_collector as trc
+from .mocks.mock_client import MockTritonClientMethods
+from .mocks.mock_perf_analyzer import MockPerfAnalyzerMethods
+from .mocks.mock_psutil import MockPSUtil
+from .mocks.mock_server_local import MockServerLocalMethods
+
+# Test Parameters
+MODEL_LOCAL_PATH = "/model_analyzer/models"
+MODEL_REPOSITORY_PATH = "/model_analyzer/models"
+PERF_BIN_PATH = "perf_analyzer"
+TRITON_LOCAL_BIN_PATH = "test_path"
+TEST_MODEL_NAME = "test_model"
+TEST_CONCURRENCY_RANGE = "1:16:2"
+CONFIG_TEST_ARG = "sync"
+TEST_GRPC_URL = "test_hostname:test_port"
+
+
+class TestPerfAnalyzerMethods(trc.TestResultCollector):
+    def setUp(self):
+        # Mocks
+        self.server_local_mock = MockServerLocalMethods()
+        self.perf_mock = MockPerfAnalyzerMethods()
+        self.client_mock = MockTritonClientMethods()
+        self.mock_psutil = MockPSUtil()
+        self.mock_psutil.start()
+        self.server_local_mock.start()
+        self.perf_mock.start()
+        self.client_mock.start()
+
+        # PerfAnalyzer config for all tests
+        self.config = PerfAnalyzerConfig()
+        self.config["model-name"] = TEST_MODEL_NAME
+        self.config["measurement-interval"] = 1000
+        self.config["measurement-request-count"] = 50
+
+        self.run_config = RunConfig({})
+        self.run_config.add_model_run_config(
+            ModelRunConfig("fake_name", MagicMock(), self.config)
+        )
+
+        self.gpus = [GPUDevice("TEST_DEVICE_NAME", 0, "TEST_PCI_BUS_ID", "TEST_UUID")]
+
+        # Triton Server
+        self.server = None
+        self.client = None
+
+    def test_perf_analyzer_config(self):
+        # Check config initializations
+        self.assertIsNone(
+            self.config[CONFIG_TEST_ARG],
+            msg="Server config had unexpected initial" f" value for {CONFIG_TEST_ARG}",
+        )
+
+        # Set value
+        self.config[CONFIG_TEST_ARG] = True
+
+        # Test get again
+        self.assertTrue(
+            self.config[CONFIG_TEST_ARG], msg=f"{CONFIG_TEST_ARG} was not set"
+        )
+
+        # Try to set an unsupported config argument, expect failure
+        with self.assertRaises(
+            TritonModelAnalyzerException,
+            msg="Expected exception on trying to set"
+            "unsupported argument in perf_analyzer"
+            "config",
+        ):
+            self.config["dummy"] = 1
+
+        # set and get value for each subtype of arguments
+        self.config["model-name"] = TEST_MODEL_NAME
+        self.assertEqual(self.config["model-name"], TEST_MODEL_NAME)
+
+        self.config["concurrency-range"] = TEST_CONCURRENCY_RANGE
+        self.assertEqual(self.config["concurrency-range"], TEST_CONCURRENCY_RANGE)
+
+        self.config["extra-verbose"] = True
+        self.assertTrue(self.config["extra-verbose"])
+
+    def test_perf_analyzer_boolean_args(self):
+        """Test that only positive boolean args get added"""
+        expected_cli_str = "-m test_model --measurement-interval=1000 --binary-search --measurement-request-count=50"
+
+        self.config["async"] = "False"
+        self.config["binary-search"] = "True"
+
+        self.assertEqual(self.config.to_cli_string(), expected_cli_str)
+
+    def test_perf_analyzer_additive_args(self):
+        shape = ["name1:1,2,3", "name2:4,5,6"]
+        expected_cli_str = "-m test_model --measurement-interval=1000 --shape=name1:1,2,3 --shape=name2:4,5,6 --measurement-request-count=50"
+
+        self.config["shape"] = shape[:]
+
+        self.assertEqual(self.config["shape"], shape)
+        self.assertEqual(self.config.to_cli_string(), expected_cli_str)
+
+        shape = "name1:1,2,3"
+        expected_cli_str = "-m test_model --measurement-interval=1000 --shape=name1:1,2,3 --measurement-request-count=50"
+        self.config["shape"] = shape
+
+        self.assertEqual(self.config.to_cli_string(), expected_cli_str)
+
+        shape = 5
+        self.config["shape"] = shape
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            self.config.to_cli_string()
+
+    def test_perf_analyzer_ssl_args(self):
+        """
+        Verify that the generated cli string passed to PA matches our expected output.
+        """
+        ssl_grpc_use_ssl = "True"
+        ssl_grpc_root_certifications_file = "a"
+        ssl_grpc_private_key_file = "b"
+        ssl_grpc_certificate_chain_file = "c"
+        ssl_https_verify_peer = 1
+        ssl_https_verify_host = 2
+        ssl_https_ca_certificates_file = "d"
+        ssl_https_client_certificate_type = "e"
+        ssl_https_client_certificate_file = "f"
+        ssl_https_private_key_type = "g"
+        ssl_https_private_key_file = "h"
+
+        expected_cli_str = (
+            f"-m test_model --measurement-interval=1000 --measurement-request-count=50 --ssl-grpc-use-ssl "
+            f"--ssl-grpc-root-certifications-file=a --ssl-grpc-private-key-file=b --ssl-grpc-certificate-chain-file=c "
+            f"--ssl-https-verify-peer=1 --ssl-https-verify-host=2 --ssl-https-ca-certificates-file=d --ssl-https-client-certificate-type=e "
+            f"--ssl-https-client-certificate-file=f --ssl-https-private-key-type=g --ssl-https-private-key-file=h"
+        )
+
+        self.config["ssl-grpc-use-ssl"] = ssl_grpc_use_ssl
+        self.config[
+            "ssl-grpc-root-certifications-file"
+        ] = ssl_grpc_root_certifications_file
+        self.config["ssl-grpc-private-key-file"] = ssl_grpc_private_key_file
+        self.config["ssl-grpc-certificate-chain-file"] = ssl_grpc_certificate_chain_file
+        self.config["ssl-https-verify-peer"] = ssl_https_verify_peer
+        self.config["ssl-https-verify-host"] = ssl_https_verify_host
+        self.config["ssl-https-ca-certificates-file"] = ssl_https_ca_certificates_file
+        self.config[
+            "ssl-https-client-certificate-type"
+        ] = ssl_https_client_certificate_type
+        self.config[
+            "ssl-https-client-certificate-file"
+        ] = ssl_https_client_certificate_file
+        self.config["ssl-https-private-key-type"] = ssl_https_private_key_type
+        self.config["ssl-https-private-key-file"] = ssl_https_private_key_file
+
+        self.assertEqual(self.config["ssl-grpc-use-ssl"], ssl_grpc_use_ssl)
+        self.assertEqual(
+            self.config["ssl-grpc-root-certifications-file"],
+            ssl_grpc_root_certifications_file,
+        )
+        self.assertEqual(
+            self.config["ssl-grpc-private-key-file"], ssl_grpc_private_key_file
+        )
+        self.assertEqual(
+            self.config["ssl-grpc-certificate-chain-file"],
+            ssl_grpc_certificate_chain_file,
+        )
+        self.assertEqual(self.config["ssl-https-verify-peer"], ssl_https_verify_peer)
+        self.assertEqual(self.config["ssl-https-verify-host"], ssl_https_verify_host)
+        self.assertEqual(
+            self.config["ssl-https-ca-certificates-file"],
+            ssl_https_ca_certificates_file,
+        )
+        self.assertEqual(
+            self.config["ssl-https-client-certificate-type"],
+            ssl_https_client_certificate_type,
+        )
+        self.assertEqual(
+            self.config["ssl-https-client-certificate-file"],
+            ssl_https_client_certificate_file,
+        )
+        self.assertEqual(
+            self.config["ssl-https-private-key-type"], ssl_https_private_key_type
+        )
+        self.assertEqual(
+            self.config["ssl-https-private-key-file"], ssl_https_private_key_file
+        )
+
+        self.assertEqual(self.config.to_cli_string(), expected_cli_str)
+
+        # Set ssl-grpc-use-ssl to False should remove it from the cli string
+        ssl_grpc_use_ssl = "False"
+        self.config["ssl-grpc-use-ssl"] = ssl_grpc_use_ssl
+        self.assertEqual(self.config["ssl-grpc-use-ssl"], ssl_grpc_use_ssl)
+        expected_cli_str = (
+            f"-m test_model --measurement-interval=1000 --measurement-request-count=50 "
+            f"--ssl-grpc-root-certifications-file=a --ssl-grpc-private-key-file=b --ssl-grpc-certificate-chain-file=c "
+            f"--ssl-https-verify-peer=1 --ssl-https-verify-host=2 --ssl-https-ca-certificates-file=d --ssl-https-client-certificate-type=e "
+            f"--ssl-https-client-certificate-file=f --ssl-https-private-key-type=g --ssl-https-private-key-file=h"
+        )
+        self.assertEqual(self.config.to_cli_string(), expected_cli_str)
+
+    def test_run(self):
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+
+        # Create server, client, PerfAnalyzer, and wait for server ready
+        self.server = TritonServerFactory.create_server_local(
+            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=self.gpus
+        )
+
+        perf_analyzer = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=self.run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+        self.client = TritonClientFactory.create_grpc_client(server_url=TEST_GRPC_URL)
+        self.server.start()
+        self.client.wait_for_server_ready(num_retries=1)
+
+        pa_csv_mock = """Concurrency,Inferences/Second,Client Send,Network+Server Send/Recv,Server Queue,Server Compute Input,Server Compute Infer,Server Compute Output,"""
+        pa_csv_mock += """Client Recv,p50 latency,p90 latency,p95 latency,p99 latency,Avg latency,request/response,response wait,"""
+        pa_csv_mock += """Avg GPU Utilization,Avg GPU Power Usage,Max GPU Memory Usage,Total GPU Memory\n"""
+        pa_csv_mock += """1,46.8,2,187,18,34,65,16,1,4600,4700,4800,4900,5000,3,314,"""
+        pa_csv_mock += """GPU-aaf4fea0:0.809;GPU-aaf4fea1:0.901;GPU-aaf4fea2:0.745;,GPU-aaf4fea0:91.2;GPU-aaf4fea1:100;,GPU-aaf4fea0:1000000000;GPU-aaf4fea1:2000000000,GPU-aaf4fea0:1500000000;GPU-aaf4fea2:3000000000"""
+
+        # Test avg latency parsing. GPU metric is ignored for get_perf_records()
+        perf_metrics = [PerfLatencyAvg, GPUUtilization]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 5)
+
+        # Test p90 latency parsing
+        perf_metrics = [PerfLatencyP90]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 4.7)
+
+        # Test p95 latency parsing
+        perf_metrics = [PerfLatencyP95]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 4.8)
+
+        # Test p99 latency parsing
+        perf_metrics = [PerfLatencyP99]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 4.9)
+
+        # Test throughput parsing
+        perf_metrics = [PerfThroughput]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 46.8)
+
+        # Test client response wait
+        perf_metrics = [PerfClientResponseWait]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 0.314)
+
+        # Test server queue
+        perf_metrics = [PerfServerQueue]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 0.018)
+
+        # Test server compute infer
+        perf_metrics = [PerfServerComputeInfer]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 0.065)
+
+        # Test server compute input
+        perf_metrics = [PerfServerComputeInput]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 0.034)
+
+        # Test server compute infer
+        perf_metrics = [PerfServerComputeOutput]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        records = perf_analyzer.get_perf_records()
+        self.assertEqual(len(records[TEST_MODEL_NAME]), 1)
+        self.assertEqual(records[TEST_MODEL_NAME][0].value(), 0.016)
+
+        # Test Avg GPU Utilizations. Perf metric is ignored for get_gpu_records()
+        gpu_metrics = [GPUUtilization, PerfLatencyAvg]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(gpu_metrics)
+
+        records = perf_analyzer.get_gpu_records()
+        self.assertEqual(len(records), 3)
+        self.assertEqual(records[0].device_uuid(), "GPU-aaf4fea0")
+        self.assertEqual(records[0].value(), 80.9)
+        self.assertEqual(records[1].device_uuid(), "GPU-aaf4fea1")
+        self.assertEqual(records[1].value(), 90.1)
+        self.assertEqual(records[2].device_uuid(), "GPU-aaf4fea2")
+        self.assertEqual(records[2].value(), 74.5)
+
+        # Test GPU Power Usage
+        gpu_metrics = [GPUPowerUsage]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(gpu_metrics)
+
+        records = perf_analyzer.get_gpu_records()
+        self.assertEqual(len(records), 2)
+        self.assertEqual(records[0].device_uuid(), "GPU-aaf4fea0")
+        self.assertEqual(records[0].value(), 91.2)
+        self.assertEqual(records[1].device_uuid(), "GPU-aaf4fea1")
+        self.assertEqual(records[1].value(), 100)
+
+        # Test GPU Memory Usage
+        gpu_metrics = [GPUUsedMemory]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(gpu_metrics)
+
+        records = perf_analyzer.get_gpu_records()
+        self.assertEqual(len(records), 2)
+        self.assertEqual(records[0].device_uuid(), "GPU-aaf4fea0")
+        self.assertEqual(records[0].value(), 1000)
+        self.assertEqual(records[1].device_uuid(), "GPU-aaf4fea1")
+        self.assertEqual(records[1].value(), 2000)
+
+        # Test Free GPU Memory (Must be measured with GPUUsedMemory)
+        # GPU a0 has 1500 total memory and 1000 used memory, so free == 500
+        # GPU a1 has no value reported for total, so it is ignored
+        # GPU a2 has no value reported for used, so it is ignored
+        gpu_metrics = [GPUFreeMemory, GPUUsedMemory]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(gpu_metrics)
+
+        records = perf_analyzer.get_gpu_records()
+        self.assertEqual(len(records), 3)
+        self.assertEqual(type(records[0]), GPUUsedMemory)
+        self.assertEqual(type(records[1]), GPUUsedMemory)
+        self.assertEqual(type(records[2]), GPUFreeMemory)
+        self.assertEqual(records[2].device_uuid(), "GPU-aaf4fea0")
+        self.assertEqual(records[2].value(), 1500 - 1000)
+
+        # # Test parsing for subset
+        perf_metrics = [
+            PerfThroughput,
+            PerfLatencyAvg,
+            PerfLatencyP90,
+            PerfLatencyP95,
+            PerfLatencyP99,
+            GPUUtilization,
+            GPUPowerUsage,
+        ]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(perf_metrics)
+
+        perf_records = perf_analyzer.get_perf_records()
+        gpu_records = perf_analyzer.get_gpu_records()
+
+        self.assertEqual(len(perf_records[TEST_MODEL_NAME]), 5)
+        # GPUPowerUsage has 2 devices and GPUUtilization has 3
+        self.assertEqual(len(gpu_records), 5)
+
+        # Test no exceptions are raised when nothing can be parsed
+        pa_csv_empty = ""
+        perf_metrics = [
+            PerfThroughput,
+            PerfClientSendRecv,
+            PerfClientResponseWait,
+            PerfServerQueue,
+            PerfServerComputeInfer,
+            PerfServerComputeInput,
+            PerfServerComputeOutput,
+            GPUFreeMemory,
+        ]
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_empty),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            self.assertFalse(perf_analyzer.run(perf_metrics))
+
+        # Test case where PA returns blank values for some GPU metrics
+        pa_csv_mock = """Concurrency,Inferences/Second,Client Send,Network+Server Send/Recv,Server Queue,Server Compute Input,Server Compute Infer,Server Compute Output,"""
+        pa_csv_mock += """Client Recv,p50 latency,p90 latency,p95 latency,p99 latency,Avg latency,request/response,response wait,"""
+        pa_csv_mock += """Avg GPU Utilization,Avg GPU Power Usage,Max GPU Memory Usage,Total GPU Memory\n"""
+        pa_csv_mock += """1,46.8,2,187,18,34,65,16,1,4600,4700,4800,4900,5000,3,314,"""
+        pa_csv_mock += """,,,7:1500"""
+
+        # Test Max GPU Memory
+        gpu_metrics = [GPUUsedMemory]
+
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            perf_analyzer.run(gpu_metrics)
+
+        records = perf_analyzer.get_gpu_records()
+        self.assertEqual(len(records), 0)
+
+        # Test exception handling
+        self.perf_mock.set_perf_analyzer_return_code(1)
+        with patch(
+            "model_analyzer.perf_analyzer.perf_analyzer.open",
+            mock_open(read_data=pa_csv_mock),
+        ), patch("model_analyzer.perf_analyzer.perf_analyzer.os.remove"):
+            self.assertTrue(perf_analyzer.run(perf_metrics))
+        self.server.stop()
+
+    def test_measurement_interval_increase(self):
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+
+        # Create server, client, PerfAnalyzer, and wait for server ready
+        self.server = TritonServerFactory.create_server_local(
+            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=self.gpus
+        )
+        perf_analyzer_config = PerfAnalyzerConfig()
+        perf_analyzer_config["model-name"] = TEST_MODEL_NAME
+        perf_analyzer_config["concurrency-range"] = TEST_CONCURRENCY_RANGE
+        perf_analyzer_config["measurement-mode"] = "time_windows"
+        perf_analyzer = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=self.run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+        self.client = TritonClientFactory.create_grpc_client(server_url=TEST_GRPC_URL)
+        self.server.start()
+
+        # Test failure to stabilize for measurement windows
+        self.client.wait_for_server_ready(num_retries=1)
+        test_stabilize_output = "Please use a larger time window"
+        self.perf_mock.set_perf_analyzer_result_string(test_stabilize_output)
+        self.perf_mock.set_perf_analyzer_return_code(1)
+        perf_metrics = [PerfThroughput, PerfLatencyP99]
+        perf_analyzer.run(perf_metrics)
+        self.assertEqual(self.perf_mock.get_perf_analyzer_popen_call_count(), 10)
+
+    def test_measurement_request_count_increase(self):
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+
+        # Create server, client, PerfAnalyzer, and wait for server ready
+        self.server = TritonServerFactory.create_server_local(
+            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=self.gpus
+        )
+        perf_analyzer = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=self.run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+        self.client = TritonClientFactory.create_grpc_client(server_url=TEST_GRPC_URL)
+        self.server.start()
+
+        # Test the timeout for count mode
+        self.client.wait_for_server_ready(num_retries=1)
+        test_both_output = "Please use a larger time window"
+        self.perf_mock.set_perf_analyzer_result_string(test_both_output)
+        self.perf_mock.set_perf_analyzer_return_code(1)
+        perf_metrics = [PerfThroughput, PerfLatencyP99]
+        perf_analyzer.run(perf_metrics)
+        self.assertEqual(self.perf_mock.get_perf_analyzer_popen_call_count(), 10)
+
+    def test_is_multi_model(self):
+        """
+        Test the functionality of the _is_multi_model() function
+
+        If the provided run_config only has one ModelRunConfig, then is_multi_model is false. Any
+        more than one ModelRunConfig and it should return true
+        """
+        run_config = RunConfig({})
+        run_config.add_model_run_config(
+            ModelRunConfig(MagicMock(), MagicMock(), MagicMock())
+        )
+
+        pa1 = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+        self.assertFalse(pa1._is_multi_model())
+
+        run_config.add_model_run_config(
+            ModelRunConfig(MagicMock(), MagicMock(), MagicMock())
+        )
+        pa2 = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+        self.assertTrue(pa2._is_multi_model())
+
+    def test_get_cmd_single_model(self):
+        """
+        Test the functionality of _get_cmd() for single model
+        """
+        pa = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=self.run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+
+        expected_cmd = [
+            "perf_analyzer",
+            "-m",
+            "test_model",
+            "--measurement-interval",
+            "1000",
+            "--measurement-request-count",
+            "50",
+        ]
+        self.assertEqual(pa._get_cmd(), expected_cmd)
+
+    def test_get_cmd_multi_model(self):
+        """
+        Test the functionality of _get_cmd() for multi model
+        """
+        pac1 = PerfAnalyzerConfig()
+        pac1["model-name"] = "MultiModel1"
+        pac1["measurement-interval"] = 1000
+        pac1["measurement-request-count"] = 50
+
+        pac2 = PerfAnalyzerConfig()
+        pac2["model-name"] = "MultiModel2"
+        pac2["batch-size"] = 16
+        pac2["concurrency-range"] = 1024
+
+        run_config = RunConfig({})
+        run_config.add_model_run_config(ModelRunConfig(MagicMock(), MagicMock(), pac1))
+        run_config.add_model_run_config(ModelRunConfig(MagicMock(), MagicMock(), pac2))
+
+        pa = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+
+        # yapf: disable
+        expected_cmd = [
+            'mpiexec', '--allow-run-as-root', '--tag-output',
+            '-n', '1', 'perf_analyzer', '--enable-mpi',
+                '-m', 'MultiModel1',
+                '--measurement-interval', '1000',
+                '--measurement-request-count', '50',
+            ':', '-n', '1', 'perf_analyzer', '--enable-mpi',
+                '-m', 'MultiModel2',
+                '-b', '16',
+                '--concurrency-range', '1024'
+        ]
+        # yapf: enable
+
+        self.assertEqual(pa._get_cmd(), expected_cmd)
+
+    def test_split_output_per_rank_for_single_model(self):
+        """
+        Test functionality of _get_output_per_rank() for single-model
+        """
+
+        output = """
+[1,0]<stdout>:*** Measurement Settings ***
+[1,1]<stdout>:*** Measurement Settings ***
+[1,1]<stdout>:Request concurrency: 2
+[1,0]<stdout>:Request concurrency: 1
+"""
+
+        # Expect no change for single-model
+        expected_result = [output]
+
+        run_config = RunConfig({})
+        run_config.add_model_run_config(
+            ModelRunConfig(MagicMock(), MagicMock(), MagicMock())
+        )
+
+        pa = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+
+        pa._output = output
+        result = pa._split_output_per_rank()
+        self.assertEqual(result, expected_result)
+
+    def test_split_output_per_rank_for_multi_model(self):
+        """
+        Test functionality of _get_output_per_rank() for multi-model
+        """
+
+        output = """
+[1,0]<stdout>:*** Measurement Settings ***
+[1,1]<stdout>:*** Measurement Settings ***
+[1,1]<stdout>:Request concurrency: 2
+[1,0]<stdout>:Request concurrency: 1
+"""
+
+        expected_result = [
+            """[1,0]<stdout>:*** Measurement Settings ***
+[1,0]<stdout>:Request concurrency: 1
+""",
+            """[1,1]<stdout>:*** Measurement Settings ***
+[1,1]<stdout>:Request concurrency: 2
+""",
+        ]
+
+        run_config = RunConfig({})
+        run_config.add_model_run_config(
+            ModelRunConfig(MagicMock(), MagicMock(), MagicMock())
+        )
+        run_config.add_model_run_config(
+            ModelRunConfig(MagicMock(), MagicMock(), MagicMock())
+        )
+
+        pa = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+
+        pa._output = output
+        result = pa._split_output_per_rank()
+        self.assertEqual(result, expected_result)
+
+    def test_auto_adjust_parameters(self):
+        """
+        Test function _auto_adjust_parameters()
+
+        In the case below:
+            - Model0 fails to be stable in time_windows and should have measurement-interval increased
+            - Model1 is fine and should have no changes
+            - Model2 fails to be stable in count_windows and should have measurement-request-count increased
+        """
+        pac0 = PerfAnalyzerConfig()
+        pac0["model-name"] = "MultiModel0"
+        pac0["measurement-mode"] = "time_windows"
+
+        pac1 = PerfAnalyzerConfig()
+        pac1["model-name"] = "MultiModel1"
+        pac1["measurement-mode"] = "time_windows"
+
+        pac2 = PerfAnalyzerConfig()
+        pac2["model-name"] = "MultiModel2"
+        pac2["measurement-mode"] = "count_windows"
+
+        run_config = RunConfig({})
+        run_config.add_model_run_config(ModelRunConfig(MagicMock(), MagicMock(), pac0))
+        run_config.add_model_run_config(ModelRunConfig(MagicMock(), MagicMock(), pac1))
+        run_config.add_model_run_config(ModelRunConfig(MagicMock(), MagicMock(), pac2))
+
+        pa = PerfAnalyzer(
+            path=PERF_BIN_PATH,
+            config=run_config,
+            max_retries=10,
+            timeout=100,
+            max_cpu_util=50,
+        )
+
+        pa._output = """
+[1,0]<stdout>:*** Measurement Settings ***
+[1,1]<stdout>:*** Measurement Settings ***
+[1,2]<stdout>:*** Measurement Settings ***
+[1,2]<stdout>:Failed to obtain stable measurement
+[1,0]<stdout>:Failed to obtain stable measurement
+[1,1]<stdout>:Success
+        """
+
+        pa._auto_adjust_parameters(MagicMock())
+
+        expected_measurement_interval = (
+            PERF_ANALYZER_MEASUREMENT_WINDOW + MEASUREMENT_WINDOW_STEP
+        )
+        expected_request_count = (
+            PERF_ANALYZER_MINIMUM_REQUEST_COUNT + MEASUREMENT_REQUEST_COUNT_STEP
+        )
+
+        self.assertEqual(
+            expected_measurement_interval,
+            pa._config.model_run_configs()[0].perf_config()["measurement-interval"],
+        )
+        self.assertEqual(
+            None,
+            pa._config.model_run_configs()[0].perf_config()[
+                "measurement-request-count"
+            ],
+        )
+        self.assertEqual(
+            None,
+            pa._config.model_run_configs()[1].perf_config()["measurement-interval"],
+        )
+        self.assertEqual(
+            None,
+            pa._config.model_run_configs()[1].perf_config()[
+                "measurement-request-count"
+            ],
+        )
+        self.assertEqual(
+            None,
+            pa._config.model_run_configs()[2].perf_config()["measurement-interval"],
+        )
+        self.assertEqual(
+            expected_request_count,
+            pa._config.model_run_configs()[2].perf_config()[
+                "measurement-request-count"
+            ],
+        )
+
+    def tearDown(self):
+        # In case test raises exception
+        if self.server is not None:
+            self.server.stop()
+
+        # Stop mocking
+        self.server_local_mock.stop()
+        self.perf_mock.stop()
+        self.client_mock.stop()
+        self.mock_psutil.stop()
+        patch.stopall()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_perf_analyzer_config_generator.py b/src/c++/perf_analyzer/tests/test_perf_analyzer_config_generator.py
new file mode 100755
index 000000000..e9852356e
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_perf_analyzer_config_generator.py
@@ -0,0 +1,842 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.generate.generator_utils import GeneratorUtils as utils
+from model_analyzer.config.generate.perf_analyzer_config_generator import (
+    PerfAnalyzerConfigGenerator,
+)
+from model_analyzer.config.input.config_defaults import (
+    DEFAULT_RUN_CONFIG_MAX_CONCURRENCY,
+    DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE,
+    DEFAULT_RUN_CONFIG_MIN_REQUEST_RATE,
+)
+from tests.common.test_utils import (
+    construct_perf_analyzer_config,
+    construct_run_config_measurement,
+    evaluate_mock_config,
+)
+
+from .common import test_result_collector as trc
+from .mocks.mock_os import MockOSMethods
+
+
+class TestPerfAnalyzerConfigGenerator(trc.TestResultCollector):
+    def __init__(self, methodname):
+        super().__init__(methodname)
+        self._perf_throughput = 1
+
+    def test_set_last_results(self):
+        """
+        Test set_last_results() with multi model
+
+        Confirm that set_last_results will properly choose the measurement with
+        the highest total throughput
+        """
+        measurement1 = self.make_multi_model_measurement(
+            ["modelA", "modelB"], [{"perf_throughput": 1}, {"perf_throughput": 2}]
+        )
+
+        measurement2 = self.make_multi_model_measurement(
+            ["modelA", "modelB"], [{"perf_throughput": 7}, {"perf_throughput": 7}]
+        )
+
+        measurement3 = self.make_multi_model_measurement(
+            ["modelA", "modelB"], [{"perf_throughput": 10}, {"perf_throughput": 2}]
+        )
+
+        pacg = PerfAnalyzerConfigGenerator(
+            MagicMock(), MagicMock(), MagicMock(), MagicMock(), early_exit_enable=False
+        )
+
+        pacg.set_last_results([measurement1, measurement2, measurement3])
+        self.assertEqual(pacg._last_results[0], measurement2)
+
+    def test_default(self):
+        """
+        Test Default:
+            - No CLI options specified
+
+        Default (1) value will be used for batch size
+        and log2(DEFAULT_RUN_CONFIG_MAX_CONCURRENCY)+1 configs
+        will be generated by the auto-search
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        concurrencies = utils.generate_doubled_list(
+            1, DEFAULT_RUN_CONFIG_MAX_CONCURRENCY
+        )
+        expected_configs = [
+            construct_perf_analyzer_config(concurrency=c) for c in concurrencies
+        ]
+
+        self._run_and_test_perf_analyzer_config_generator(yaml_str, expected_configs)
+
+    def test_search_disabled(self):
+        """
+        Test Search Disabled:
+            - Run Config Search disabled
+
+        Default (1) value will be used for batch size
+        and concurrency will be set to 1
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        expected_configs = [construct_perf_analyzer_config()]
+
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, "--run-config-search-disable"
+        )
+
+    def test_c_api(self):
+        """
+        Test C_API:
+            - Launch mode is C_API
+
+        Default (1) values will be used for batch size/concurrency
+        and only one config will be generated
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        concurrencies = utils.generate_doubled_list(
+            1, DEFAULT_RUN_CONFIG_MAX_CONCURRENCY
+        )
+        expected_configs = [
+            construct_perf_analyzer_config(concurrency=c, launch_mode="c_api")
+            for c in concurrencies
+        ]
+
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, "--triton-launch-mode=c_api"
+        )
+
+    def test_http(self):
+        """
+        Test HTTP:
+            - Client protocol is HTTP
+
+        Default (1) values will be used for batch size/concurrency
+        and only one config will be generated
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        concurrencies = utils.generate_doubled_list(
+            1, DEFAULT_RUN_CONFIG_MAX_CONCURRENCY
+        )
+        expected_configs = [
+            construct_perf_analyzer_config(concurrency=c, client_protocol="http")
+            for c in concurrencies
+        ]
+
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, "--client-protocol=http"
+        )
+
+    def test_batch_size_search_disabled(self):
+        """
+        Test Batch Size Search Disabled:
+            - Schmoo batch sizes
+            - Run Config Search disabled
+
+        Batch sizes: 1,2,4
+        Default (1) value will be used concurrency
+        and 3 configs will be generated
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        batch_sizes = [1, 2, 4]
+        expected_configs = [
+            construct_perf_analyzer_config(batch_size=b) for b in batch_sizes
+        ]
+
+        pa_cli_args = ["-b 1,2,4", "--run-config-search-disable"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_batch_size_search_enabled(self):
+        """
+        Test Batch Size Search Enabled:
+            - Schmoo batch sizes
+            - Run Config Search enabled
+
+        Batch sizes: 1,2,4
+        Concurrency: log2(DEFAULT_RUN_CONFIG_MAX_CONCURRENCY)+1
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        batch_sizes = [1, 2, 4]
+        concurrencies = utils.generate_doubled_list(
+            1, DEFAULT_RUN_CONFIG_MAX_CONCURRENCY
+        )
+        expected_configs = [
+            construct_perf_analyzer_config(batch_size=b, concurrency=c)
+            for b in batch_sizes
+            for c in concurrencies
+        ]
+
+        pa_cli_args = ["-b 1,2,4"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_concurrency(self):
+        """
+        Test Concurrency:
+            - Schmoo concurrency
+            - Test with auto-search enabled & disabled
+
+        Concurrency: 1-4
+        Default (1) value will be used for batch size
+        and 4 configs will be generated
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        concurrencies = [1, 2, 3, 4]
+        expected_configs = [
+            construct_perf_analyzer_config(concurrency=c) for c in concurrencies
+        ]
+
+        pa_cli_args = ["-c 1,2,3,4"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+        pa_cli_args = ["-c 1,2,3,4", "--run-config-search-disable"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_batch_size_and_concurrency(self):
+        """
+        Test Batch Size and Concurrency:
+            - Schmoo batch sizes and concurrency
+            - Run Config Search enabled & disabled
+
+        Batch sizes: 1,2,4
+        Concurrency: 1-4
+
+
+        12 configs will be generated
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        batch_sizes = [1, 2, 4]
+        concurrencies = [1, 2, 3, 4]
+
+        expected_configs = [
+            construct_perf_analyzer_config(batch_size=b, concurrency=c)
+            for b in batch_sizes
+            for c in concurrencies
+        ]
+
+        pa_cli_args = ["-b 1,2,4", "-c 1,2,3,4"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+        pa_cli_args = ["-b 1,2,4", "-c 1,2,3,4", "--run-config-search-disable"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_max_concurrency(self):
+        """
+        Test Max Concurrency:
+            - Change max concurrency to non-default value
+
+        Max Concurrency: 16
+        Default (1) value will be used for batch size
+        and 5 configs (log2(16)+1) will be generated
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        concurrencies = utils.generate_doubled_list(1, 16)
+        expected_configs = [
+            construct_perf_analyzer_config(concurrency=c) for c in concurrencies
+        ]
+
+        pa_cli_args = ["--run-config-search-max-concurrency", "16"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_min_concurrency(self):
+        """
+        Test Min Concurrency:
+            - Change min concurrency to non-default value
+
+        Min Concurrency: 5
+        2 configs [5, 10] will be generated
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        concurrencies = [5, 10]
+        expected_configs = [
+            construct_perf_analyzer_config(concurrency=c) for c in concurrencies
+        ]
+
+        pa_cli_args = [
+            "--run-config-search-min-concurrency",
+            "5",
+            "--run-config-search-max-concurrency",
+            "16",
+        ]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_request_rate_list(self):
+        """
+        Test Request Rate:
+            - Shmoo request-rate
+            - Test with auto-search enabled & disabled
+
+        Request Rate: 1,2,3,4
+        Default (1) value will be used for batch size
+        and 4 configs will be generated
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        request_rates = [1, 2, 3, 4]
+        expected_configs = [
+            construct_perf_analyzer_config(request_rate=request_rate)
+            for request_rate in request_rates
+        ]
+
+        pa_cli_args = ["--request-rate", "1,2,3,4"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+        pa_cli_args = ["--request-rate", "1,2,3,4", "--run-config-search-disable"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+              my-model:
+                parameters:
+                  request_rate: 1,2,3,4
+            """)
+        # yapf: enable
+
+        pa_cli_args = []
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_request_rate_enable(self):
+        """
+        Test Request Rate:
+            - Shmoo request rate
+            - Test with auto-search enabled & disabled
+
+        Request Rate: MIN-MAX
+        Default (1) value will be used for batch size
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        request_rates = utils.generate_doubled_list(
+            DEFAULT_RUN_CONFIG_MIN_REQUEST_RATE, DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE
+        )
+        expected_configs = [
+            construct_perf_analyzer_config(request_rate=request_rate)
+            for request_rate in request_rates
+        ]
+
+        pa_cli_args = ["--request-rate-search-enable"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_max_request_rate(self):
+        """
+        Test Max Request Rate:
+            - Change max request rate to non-default value
+
+        Max Request Rate: DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE / 2
+        Default (1) value will be used for batch size
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        request_rates = utils.generate_doubled_list(
+            DEFAULT_RUN_CONFIG_MIN_REQUEST_RATE,
+            int(DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE / 2),
+        )
+        expected_configs = [
+            construct_perf_analyzer_config(request_rate=request_rate)
+            for request_rate in request_rates
+        ]
+
+        pa_cli_args = [
+            "--run-config-search-max-request-rate",
+            f"{int(DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE / 2)}",
+        ]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_min_request_rate(self):
+        """
+        Test Min Request Rate:
+            - Change max request rate to non-default value
+
+        Min Request Rate: DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE * 2
+        Default (1) value will be used for batch size
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        request_rates = utils.generate_doubled_list(
+            DEFAULT_RUN_CONFIG_MIN_REQUEST_RATE * 2,
+            int(DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE),
+        )
+        expected_configs = [
+            construct_perf_analyzer_config(request_rate=request_rate)
+            for request_rate in request_rates
+        ]
+
+        pa_cli_args = [
+            "--run-config-search-min-request-rate",
+            f"{int(DEFAULT_RUN_CONFIG_MIN_REQUEST_RATE * 2)}",
+        ]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args
+        )
+
+    def test_perf_analyzer_flags(self):
+        """
+        Test Perf Analyzer Flags:
+            - No CLI options specified
+            - Percentile (PA flag) set in model's YAML
+
+        Default (1) value will be used for batch size
+        and log2(DEFAULT_RUN_CONFIG_MAX_CONCURRENCY)+1 configs
+        will be generated by the auto-search
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model:
+                    perf_analyzer_flags:
+                        percentile: 96
+            """)
+        # yapf: enable
+
+        concurrencies = utils.generate_doubled_list(
+            1, DEFAULT_RUN_CONFIG_MAX_CONCURRENCY
+        )
+        expected_configs = [
+            construct_perf_analyzer_config(
+                concurrency=c, perf_analyzer_flags={"percentile": "96"}
+            )
+            for c in concurrencies
+        ]
+
+        self._run_and_test_perf_analyzer_config_generator(yaml_str, expected_configs)
+
+    def test_perf_analyzer_config_ssl_options(self):
+        """
+        Test Perf Analyzer SSL options:
+            - No CLI options specified
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model:
+                    perf_analyzer_flags:
+                        ssl-grpc-root-certifications-file: a
+                        ssl-grpc-private-key-file: b
+                        ssl-grpc-certificate-chain-file: c
+                        ssl-https-verify-peer: 1
+                        ssl-https-verify-host: 2
+                        ssl-https-ca-certificates-file: d
+                        ssl-https-client-certificate-type: e
+                        ssl-https-client-certificate-file: f
+                        ssl-https-private-key-type: g
+                        ssl-https-private-key-file: h
+            """)
+        # yapf: enable
+
+        concurrencies = utils.generate_doubled_list(
+            1, DEFAULT_RUN_CONFIG_MAX_CONCURRENCY
+        )
+        expected_configs = [
+            construct_perf_analyzer_config(
+                concurrency=c,
+                perf_analyzer_flags={
+                    "ssl-grpc-root-certifications-file": "a",
+                    "ssl-grpc-private-key-file": "b",
+                    "ssl-grpc-certificate-chain-file": "c",
+                    "ssl-https-verify-peer": "1",
+                    "ssl-https-verify-host": "2",
+                    "ssl-https-ca-certificates-file": "d",
+                    "ssl-https-client-certificate-type": "e",
+                    "ssl-https-client-certificate-file": "f",
+                    "ssl-https-private-key-type": "g",
+                    "ssl-https-private-key-file": "h",
+                },
+            )
+            for c in concurrencies
+        ]
+
+        self._run_and_test_perf_analyzer_config_generator(yaml_str, expected_configs)
+
+    def test_early_exit_on_no_plateau(self):
+        """
+        Test if early_exit is true but the throughput is still increasing, we
+        do not early exit
+        """
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        concurrencies = utils.generate_doubled_list(1, 64)
+        expected_configs = [
+            construct_perf_analyzer_config(concurrency=c) for c in concurrencies
+        ]
+
+        pa_cli_args = ["--run-config-search-max-concurrency", "64"]
+        self._run_and_test_perf_analyzer_config_generator(
+            yaml_str, expected_configs, pa_cli_args, early_exit=True
+        )
+
+    def test_early_exit_on_yes_plateau(self):
+        """
+        Test if early_exit is true and the throughput plateaus, we do early exit
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        concurrencies = utils.generate_doubled_list(1, 32)
+        expected_configs = [
+            construct_perf_analyzer_config(concurrency=c) for c in concurrencies
+        ]
+
+        pa_cli_args = ["--run-config-search-max-concurrency", "64"]
+        with patch.object(
+            TestPerfAnalyzerConfigGenerator, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = [1, 2, 4, 4, 4, 4, 4]
+            self._run_and_test_perf_analyzer_config_generator(
+                yaml_str, expected_configs, pa_cli_args, early_exit=True
+            )
+
+    def test_early_exit_off_yes_plateau(self):
+        """
+        Test if early_exit is off and the throughput plateaus, we do not early exit
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        concurrencies = utils.generate_doubled_list(1, 64)
+        expected_configs = [
+            construct_perf_analyzer_config(concurrency=c) for c in concurrencies
+        ]
+
+        pa_cli_args = ["--run-config-search-max-concurrency", "64"]
+        with patch.object(
+            TestPerfAnalyzerConfigGenerator, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = [1, 2, 4, 4, 4, 4, 4]
+            self._run_and_test_perf_analyzer_config_generator(
+                yaml_str, expected_configs, pa_cli_args, early_exit=False
+            )
+
+    def test_throughput_gain_based_on_max(self):
+        # Expect false because no increases
+        throughput_values = [50, 40, 30, 20]
+        expected_result = False
+        self._test_throughput_gain_valid_helper(throughput_values, expected_result)
+
+        # Expect false because no increases in the last 4
+        throughput_values = [10, 20, 30, 40, 50, 40, 30, 20]
+        expected_result = False
+        self._test_throughput_gain_valid_helper(throughput_values, expected_result)
+
+        # Expect false because gain is only 5%
+        throughput_values = [50, 50, 50, 52.5]
+        expected_result = False
+        self._test_throughput_gain_valid_helper(throughput_values, expected_result)
+
+        # Expect false because gain is only 5%
+        throughput_values = [50, 35, 45, 51.5]
+        expected_result = False
+        self._test_throughput_gain_valid_helper(throughput_values, expected_result)
+
+        # Expect true because gain is more than 5%
+        throughput_values = [50, 50, 50, 52.51]
+        expected_result = True
+        self._test_throughput_gain_valid_helper(throughput_values, expected_result)
+
+        # Expect true because not enough data
+        throughput_values = [50, 10]
+        expected_result = True
+        self._test_throughput_gain_valid_helper(throughput_values, expected_result)
+
+        # Expect true because of increases
+        throughput_values = [50, 100, 200, 400]
+        expected_result = True
+        self._test_throughput_gain_valid_helper(throughput_values, expected_result)
+
+        # Expect false because no new max
+        throughput_values = [50, 10, 50, 10]
+        expected_result = False
+        self._test_throughput_gain_valid_helper(throughput_values, expected_result)
+
+    def _test_throughput_gain_valid_helper(self, throughput_values, expected_result):
+        throughputs = [
+            construct_run_config_measurement(
+                model_name=MagicMock(),
+                model_config_names=["test_model_config_name"],
+                model_specific_pa_params=MagicMock(),
+                gpu_metric_values=MagicMock(),
+                non_gpu_metric_values=[{"perf_throughput": throughput_value}],
+            )
+            for throughput_value in throughput_values
+        ]
+
+        result = PerfAnalyzerConfigGenerator.throughput_gain_valid_helper(
+            throughputs=throughputs, min_tries=4, min_gain=0.05
+        )
+
+        self.assertEqual(result, expected_result)
+
+    def _get_next_measurement(self):
+        throughput_value = self._get_next_perf_throughput_value()
+        if throughput_value is None:
+            return None
+        else:
+            return construct_run_config_measurement(
+                model_name=MagicMock(),
+                model_config_names=["test_model_config_name"],
+                model_specific_pa_params=MagicMock(),
+                gpu_metric_values=MagicMock(),
+                non_gpu_metric_values=[{"perf_throughput": throughput_value}],
+            )
+
+    def _get_next_perf_throughput_value(self):
+        self._perf_throughput *= 2
+        return self._perf_throughput
+
+    def _run_and_test_perf_analyzer_config_generator(
+        self, yaml_str, expected_configs, pa_cli_args=None, early_exit=False
+    ):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+
+        if type(pa_cli_args) == list:
+            args = args + pa_cli_args
+        elif type(pa_cli_args) == str:
+            args.append(pa_cli_args)
+
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        pacg = PerfAnalyzerConfigGenerator(
+            config,
+            config.profile_models[0].model_name(),
+            config.profile_models[0].perf_analyzer_flags(),
+            config.profile_models[0].parameters(),
+            early_exit,
+        )
+
+        perf_analyzer_configs = []
+        for perf_config in pacg.get_configs():
+            perf_analyzer_configs.append(perf_config)
+            pacg.set_last_results([self._get_next_measurement()])
+
+        self.assertEqual(len(expected_configs), len(perf_analyzer_configs))
+        for i in range(len(expected_configs)):
+            self.assertEqual(
+                expected_configs[i]._options["-m"],
+                perf_analyzer_configs[i]._options["-m"],
+            )
+            self.assertEqual(
+                expected_configs[i]._options["-b"],
+                perf_analyzer_configs[i]._options["-b"],
+            )
+            self.assertEqual(
+                expected_configs[i]._options["-i"],
+                perf_analyzer_configs[i]._options["-i"],
+            )
+            self.assertEqual(
+                expected_configs[i]._options["-u"],
+                perf_analyzer_configs[i]._options["-u"],
+            )
+
+            self.assertEqual(
+                expected_configs[i]._args["concurrency-range"],
+                perf_analyzer_configs[i]._args["concurrency-range"],
+            )
+            self.assertEqual(
+                expected_configs[i]._args["measurement-mode"],
+                perf_analyzer_configs[i]._args["measurement-mode"],
+            )
+            self.assertEqual(
+                expected_configs[i]._args["service-kind"],
+                perf_analyzer_configs[i]._args["service-kind"],
+            )
+            self.assertEqual(
+                expected_configs[i]._args["triton-server-directory"],
+                perf_analyzer_configs[i]._args["triton-server-directory"],
+            )
+            self.assertEqual(
+                expected_configs[i]._args["model-repository"],
+                perf_analyzer_configs[i]._args["model-repository"],
+            )
+
+            # Future-proofing (in case a new field gets added)
+            self.assertEqual(
+                expected_configs[i]._options, perf_analyzer_configs[i]._options
+            )
+            self.assertEqual(expected_configs[i]._args, perf_analyzer_configs[i]._args)
+            self.assertEqual(
+                expected_configs[i]._additive_args,
+                perf_analyzer_configs[i]._additive_args,
+            )
+
+    def setUp(self):
+        # Mock path validation
+        self.mock_os = MockOSMethods(
+            mock_paths=["model_analyzer.config.input.config_utils"]
+        )
+        self.mock_os.start()
+
+    def tearDown(self):
+        self.mock_os.stop()
+        patch.stopall()
+
+    def make_multi_model_measurement(self, model_names, non_gpu_metric_values):
+        return construct_run_config_measurement(
+            model_name=MagicMock(),
+            model_config_names=model_names,
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=MagicMock(),
+            non_gpu_metric_values=non_gpu_metric_values,
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_plot.py b/src/c++/perf_analyzer/tests/test_plot.py
new file mode 100755
index 000000000..c9e3d84c0
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_plot.py
@@ -0,0 +1,131 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.plots.simple_plot import SimplePlot
+
+from .common import test_result_collector as trc
+from .common.test_utils import construct_run_config_measurement
+from .mocks.mock_matplotlib import MockMatplotlibMethods
+
+
+class TestPlotMethods(trc.TestResultCollector):
+    def setUp(self):
+        # mocks
+        self.matplotlib_mock = MockMatplotlibMethods()
+        self.matplotlib_mock.start()
+
+    def test_create_plot(self):
+        # Create a plot and check for call to subplots
+        SimplePlot(
+            name="test_plot",
+            title="test_title",
+            x_axis="perf_throughput",
+            y_axis="perf_latency_p99",
+        )
+
+        self.matplotlib_mock.assert_called_subplots()
+
+    def test_add_measurement(self):
+        plot = SimplePlot(
+            name="test_plot",
+            title="test_title",
+            x_axis="perf_throughput",
+            y_axis="perf_latency_p99",
+        )
+
+        gpu_metric_values = {0: {"gpu_used_memory": 5000, "gpu_utilization": 50}}
+
+        non_gpu_metric_values = {"perf_throughput": 200, "perf_latency_p99": 8000}
+
+        objective_spec = {"perf_throughput": 10, "perf_latency_p99": 5}
+
+        measurement = construct_run_config_measurement(
+            model_name="test_model",
+            model_config_names=["test_model_config_0"],
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=gpu_metric_values,
+            non_gpu_metric_values=[non_gpu_metric_values],
+            metric_objectives=[objective_spec],
+        )
+
+        # Add above measurement
+        plot.add_run_config_measurement("test_model_label1", measurement)
+        self.assertDictEqual(
+            plot.data(), {"test_model_label1": {"x_data": [200], "y_data": [8000]}}
+        )
+
+        # Add measurement again with different label
+        plot.add_run_config_measurement("test_model_label2", measurement)
+        self.assertDictEqual(
+            plot.data(),
+            {
+                "test_model_label1": {"x_data": [200], "y_data": [8000]},
+                "test_model_label2": {"x_data": [200], "y_data": [8000]},
+            },
+        )
+
+    def test_plot_data(self):
+        plot = SimplePlot(
+            name="test_plot",
+            title="test_title",
+            x_axis="perf_throughput",
+            y_axis="perf_latency_p99",
+        )
+
+        gpu_metric_values = {0: {"gpu_used_memory": 5000, "gpu_utilization": 50}}
+
+        non_gpu_metric_values = {"perf_throughput": 200, "perf_latency_p99": 8000}
+
+        objective_spec = {"perf_throughput": 10, "perf_latency_p99": 5}
+
+        measurement = construct_run_config_measurement(
+            model_name="test_model",
+            model_config_names=["test_model_config_0"],
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=gpu_metric_values,
+            non_gpu_metric_values=[non_gpu_metric_values],
+            metric_objectives=[objective_spec],
+        )
+
+        plot.add_run_config_measurement("test_model_label", measurement)
+
+        # Call plot and assert args
+        plot.plot_data_and_constraints(constraints={})
+        self.matplotlib_mock.assert_called_plot_with_args(
+            x_data=[200], y_data=[8000], marker="o", label="test_model_label"
+        )
+
+    def test_save(self):
+        plot = SimplePlot(
+            name="test_plot",
+            title="test_title",
+            x_axis="perf_throughput",
+            y_axis="perf_latency_p99",
+        )
+
+        plot.save("test_path")
+        self.matplotlib_mock.assert_called_save_with_args("test_path/test_plot")
+
+    def tearDown(self):
+        self.matplotlib_mock.stop()
+        patch.stopall()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_plot_manager.py b/src/c++/perf_analyzer/tests/test_plot_manager.py
new file mode 100755
index 000000000..a8e5aa3e4
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_plot_manager.py
@@ -0,0 +1,196 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import json
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.plots.plot_manager import PlotManager
+from model_analyzer.result.constraint_manager import ConstraintManager
+from model_analyzer.result.result_manager import ResultManager
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+
+from .common import test_result_collector as trc
+from .common.test_utils import (
+    ROOT_DIR,
+    convert_to_bytes,
+    default_encode,
+    evaluate_mock_config,
+)
+
+
+class TestPlotManager(trc.TestResultCollector):
+    def setUp(self):
+        self.create_new_dump = False
+        self._create_single_model_result_manager()
+        self._create_multi_model_result_manager()
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_single_model_summary_plots_against_golden(self):
+        """
+        Match the summary plots against the golden versions in
+        tests/common/single-model-ckpt
+        """
+        plot_manager = PlotManager(
+            config=self._single_model_config,
+            result_manager=self._single_model_result_manager,
+            constraint_manager=MagicMock(),
+        )
+
+        plot_manager.create_summary_plots()
+
+        # Set self.create_new_dump to True for creating a new output dump to compare against.
+        # Default value is False.
+        if self.create_new_dump:
+            with open(f"{ROOT_DIR}/single-model-ckpt/plot_manager.json", "wb") as f:
+                f.write(
+                    convert_to_bytes(
+                        json.dumps(
+                            self._plot_manager_to_dict(plot_manager),
+                            default=default_encode,
+                        )
+                    )
+                )
+
+        with open(f"{ROOT_DIR}/single-model-ckpt/plot_manager.json", "rb") as f:
+            golden_plot_manager_dict = json.load(f)
+
+        plot_manager_dict = self._plot_manager_to_dict(plot_manager)
+
+        self.assertEqual(golden_plot_manager_dict, plot_manager_dict)
+
+    def test_multi_model_summary_plots_against_golden(self):
+        """
+        Match the summary plots against the golden versions in
+        tests/common/multi-model-ckpt
+        """
+        plot_manager = PlotManager(
+            config=self._multi_model_config,
+            result_manager=self._multi_model_result_manager,
+            constraint_manager=MagicMock(),
+        )
+
+        plot_manager.create_summary_plots()
+
+        # Set self.create_new_dump to True for creating a new output dump to compare against.
+        # Default value is False.
+        if self.create_new_dump:
+            with open(f"{ROOT_DIR}/multi-model-ckpt/plot_manager.json", "wb") as f:
+                f.write(
+                    convert_to_bytes(
+                        json.dumps(
+                            self._plot_manager_to_dict(plot_manager),
+                            default=default_encode,
+                        )
+                    )
+                )
+
+        with open(f"{ROOT_DIR}/multi-model-ckpt/plot_manager.json", "rb") as f:
+            golden_plot_manager_dict = json.load(f)
+
+        plot_manager_dict = self._plot_manager_to_dict(plot_manager)
+
+        self.assertEqual(golden_plot_manager_dict, plot_manager_dict)
+
+    def _create_single_model_result_manager(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "-f",
+            "config.yml",
+            "--checkpoint-directory",
+            f"{ROOT_DIR}/single-model-ckpt/",
+            "--export-path",
+            f"{ROOT_DIR}/single-model-ckpt/",
+            "--model-repository",
+            ".",
+        ]
+        yaml_str = """
+            profile_models: add_sub
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+        state_manager = AnalyzerStateManager(config=config, server=None)
+        state_manager.load_checkpoint(checkpoint_required=True)
+
+        constraint_manager = ConstraintManager(config)
+        self._single_model_result_manager = ResultManager(
+            config=config,
+            state_manager=state_manager,
+            constraint_manager=constraint_manager,
+        )
+
+        self._single_model_config = config
+
+    def _create_multi_model_result_manager(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "-f",
+            "config.yml",
+            "--checkpoint-directory",
+            f"{ROOT_DIR}/multi-model-ckpt/",
+            "--export-path",
+            f"{ROOT_DIR}/multi-model-ckpt/",
+            "--model-repository",
+            ".",
+            "--run-config-profile-models-concurrently-enable",
+            "--run-config-search-mode",
+            "quick",
+        ]
+        yaml_str = """
+            profile_models: resnet50_libtorch,vgg19_libtorch
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+        state_manager = AnalyzerStateManager(config=config, server=None)
+        state_manager.load_checkpoint(checkpoint_required=True)
+
+        constraint_manager = ConstraintManager(config)
+        self._multi_model_result_manager = ResultManager(
+            config=config,
+            state_manager=state_manager,
+            constraint_manager=constraint_manager,
+        )
+
+        self._multi_model_config = config
+
+    def _plot_manager_to_dict(self, plot_manager):
+        plot_manager_dict = {}
+        plot_manager_dict["_simple_plots"] = {}
+        for spd_key, simple_plot_dict in plot_manager._simple_plots.items():
+            plot_manager_dict["_simple_plots"][spd_key] = {}
+            for sp_key, simple_plot in simple_plot_dict.items():
+                plot_manager_dict["_simple_plots"][spd_key][
+                    sp_key
+                ] = self._simple_plot_to_dict(simple_plot)
+
+        return plot_manager_dict
+
+    def _simple_plot_to_dict(self, simple_plot):
+        simple_plot_dict = {}
+        simple_plot_dict["_name"] = simple_plot._name
+        simple_plot_dict["_title"] = simple_plot._title
+        simple_plot_dict["_x_axis"] = simple_plot._x_axis
+        simple_plot_dict["_y_axis"] = simple_plot._y_axis
+        simple_plot_dict["_monotonic"] = simple_plot._monotonic
+        simple_plot_dict["_data"] = simple_plot._data
+
+        return simple_plot_dict
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_quick_run_config_generator.py b/src/c++/perf_analyzer/tests/test_quick_run_config_generator.py
new file mode 100755
index 000000000..7eef2514f
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_quick_run_config_generator.py
@@ -0,0 +1,1359 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.generate.coordinate import Coordinate
+from model_analyzer.config.generate.model_profile_spec import ModelProfileSpec
+from model_analyzer.config.generate.model_variant_name_manager import (
+    ModelVariantNameManager,
+)
+from model_analyzer.config.generate.quick_run_config_generator import (
+    QuickRunConfigGenerator,
+)
+from model_analyzer.config.generate.run_config_generator_factory import (
+    RunConfigGeneratorFactory,
+)
+from model_analyzer.config.generate.search_config import SearchConfig
+from model_analyzer.config.generate.search_dimension import SearchDimension
+from model_analyzer.config.generate.search_dimensions import SearchDimensions
+from model_analyzer.config.input.objects.config_model_profile_spec import (
+    ConfigModelProfileSpec,
+)
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.triton.model.model_config import ModelConfig
+from tests.common.test_utils import evaluate_mock_config
+
+from .common import test_result_collector as trc
+
+
+def mock_ensemble_configs(*args, **kwargs):
+    fake_config = {
+        "name": "my-model",
+        "platform": "ensemble",
+        "ensemble_scheduling": {
+            "step": [{"model_name": "fake_model_A"}, {"model_name": "fake_model_B"}]
+        },
+        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+        "max_batch_size": 4,
+    }
+    fake_base_composing_config0 = {
+        "name": "fake_model_A",
+        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+        "max_batch_size": 4,
+        "sequence_batching": {},
+    }
+    fake_base_composing_config1 = {
+        "name": "fake_model_B",
+        "input": [{"name": "INPUT__2", "dataType": "TYPE_FP16", "dims": [32]}],
+        "max_batch_size": 8,
+    }
+
+    if args:
+        model_name = args[4]
+    else:
+        model_name = kwargs["model_name"]
+
+    if model_name == "my-model":
+        return fake_config
+    elif model_name == "fake_model_A":
+        return fake_base_composing_config0
+    elif model_name == "fake_model_B":
+        return fake_base_composing_config1
+
+    return None
+
+
+def mock_composing_ensemble_configs(*args, **kwargs):
+    fake_config = {
+        "name": "my-model",
+        "platform": "ensemble",
+        "ensemble_scheduling": {
+            "step": [{"model_name": "fake_model_A"}, {"model_name": "fake_model_B"}]
+        },
+        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+        "max_batch_size": 4,
+    }
+    fake_base_composing_config0 = {
+        "name": "fake_model_A",
+        "platform": "ensemble",
+        "ensemble_scheduling": {
+            "step": [{"model_name": "fake_model_C"}, {"model_name": "fake_model_D"}]
+        },
+        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+        "max_batch_size": 4,
+        "sequence_batching": {},
+    }
+
+    if args:
+        model_name = args[4]
+    else:
+        model_name = kwargs["model_name"]
+
+    if model_name == "my-model":
+        return fake_config
+    else:
+        return fake_base_composing_config0
+
+
+def mock_bls_configs(*args, **kwargs):
+    fake_config = {
+        "name": "my-model",
+        "platform": "bls",
+        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+        "max_batch_size": 4,
+    }
+    fake_base_composing_config0 = {
+        "name": "fake_model_A",
+        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+        "max_batch_size": 4,
+        "sequence_batching": {},
+    }
+    fake_base_composing_config1 = {
+        "name": "fake_model_B",
+        "input": [{"name": "INPUT__2", "dataType": "TYPE_FP16", "dims": [32]}],
+        "max_batch_size": 8,
+    }
+
+    if args:
+        model_name = args[4]
+    else:
+        model_name = kwargs["model_name"]
+
+    if model_name == "my-model":
+        return fake_config
+    elif model_name == "fake_model_A":
+        return fake_base_composing_config0
+    else:
+        return fake_base_composing_config1
+
+
+class TestQuickRunConfigGenerator(trc.TestResultCollector):
+    def setUp(self):
+        fake_config = {
+            "name": "fake_model_name",
+            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+            "max_batch_size": 4,
+        }
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            return_value=fake_config,
+        ):
+            self._mock_models = [
+                ModelProfileSpec(
+                    ConfigModelProfileSpec(model_name="fake_model_name"),
+                    MagicMock(),
+                    MagicMock(),
+                    MagicMock(),
+                )
+            ]
+
+        self._dims = SearchDimensions()
+        self._dims.add_dimensions(
+            0,
+            [
+                SearchDimension(
+                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+                SearchDimension(
+                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
+                ),
+                SearchDimension(
+                    "concurrency", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+            ],
+        )
+
+        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
+        config = self._create_config()
+        self._qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            self._mock_models,
+            {},
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+    def test_get_starting_coordinate(self):
+        """Test that get_starting_coordinate() works for non-zero values"""
+        # yapf: disable
+        dims = SearchDimensions()
+        dims.add_dimensions(0, [
+                SearchDimension("x", SearchDimension.DIMENSION_TYPE_EXPONENTIAL, min=2),
+                SearchDimension("y", SearchDimension.DIMENSION_TYPE_LINEAR, min=1),
+                SearchDimension("z", SearchDimension.DIMENSION_TYPE_EXPONENTIAL, min=3)
+        ])
+        sc = SearchConfig(dimensions=dims,radius=2, min_initialized=2)
+        # yapf: enable
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            MagicMock(),
+            MagicMock(),
+            self._mock_models,
+            {},
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+        self.assertEqual(qrcg._get_starting_coordinate(), Coordinate([2, 1, 3]))
+
+    def test_get_next_run_config(self):
+        """
+        Test that get_next_run_config() creates a proper RunConfig
+
+        Sets up a case where the coordinate is [5,7], which corresponds to
+          - max_batch_size = 32
+          - instance_count = 8
+          - concurrency = 32*8*2 = 512
+
+        Also
+        - dynamic batching should be on
+        - existing values from the base model config should persist if they aren't overwritten
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "/tmp",
+            "--config-file",
+            "/tmp/my_config.yml",
+        ]
+
+        yaml_str = """
+        run_config_search_mode: quick
+        profile_models:
+          fake_model_name:
+            model_config_parameters:
+              dynamic_batching: { "preferred_batch_size": [4] }
+              optimization:
+                execution_accelerators:
+                  gpu_execution_accelerator:
+                    - name: tensorrt
+                      parameters:
+                        precision_mode: FP16
+        """
+
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        fake_config = {
+            "name": "fake_model_name",
+            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+            "max_batch_size": 4,
+        }
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            return_value=fake_config,
+        ):
+            mock_models = [
+                ModelProfileSpec(
+                    config.get_config()["profile_models"].value()[0],
+                    config,
+                    MagicMock(),
+                    MagicMock(),
+                )
+            ]
+
+        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
+
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            mock_models,
+            {},
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        qrcg._coordinate_to_measure = Coordinate([5, 7])
+
+        expected_model_config = {
+            "dynamicBatching": {},
+            "instanceGroup": [
+                {
+                    "count": 8,
+                    "kind": "KIND_GPU",
+                }
+            ],
+            "maxBatchSize": 32,
+            "name": "fake_model_name",
+            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
+            "optimization": {
+                "executionAccelerators": {
+                    "gpuExecutionAccelerator": [
+                        {"name": "tensorrt", "parameters": {"precision_mode": "FP16"}}
+                    ]
+                }
+            },
+        }
+        # yapf: enable
+
+        rc = qrcg._get_next_run_config()
+
+        self.assertEqual(len(rc.model_run_configs()), 1)
+        model_config = rc.model_run_configs()[0].model_config()
+        perf_config = rc.model_run_configs()[0].perf_config()
+
+        self.assertEqual(model_config.to_dict(), expected_model_config)
+        self.assertEqual(perf_config["concurrency-range"], 512)
+        self.assertEqual(perf_config["batch-size"], 1)
+
+    def test_get_next_run_config_multi_model(self):
+        """
+        Test that get_next_run_config() creates a proper RunConfig for multi-model
+
+        Sets up a case where the coordinate is [1,2,4,5], which corresponds to
+          - model 1 max_batch_size = 2
+          - model 1 instance_count = 3
+          - model 1 concurrency = 2*3*2 = 12
+          - model 2 max_batch_size = 16
+          - model 2 instance_count = 6
+          - model 2 concurrency = 16*6*2 = 192
+
+        Also,
+        - sequence batching should be on for model 1
+        - dynamic batching should be on for model 2
+        - existing values from the base model config should persist if they aren't overwritten
+        - existing values for perf-analyzer config should persist if they aren't overwritten
+        """
+
+        # yapf: disable
+        fake_base_config1 = {
+            "name": "fake_model_name1",
+            "input": [{
+                "name": "INPUT__0",
+                "dataType": "TYPE_FP32",
+                "dims": [16]
+            }],
+            "max_batch_size": 4,
+            "sequence_batching": {}
+        }
+        fake_base_config2 = {
+            "name": "fake_model_name2",
+            "input": [{
+                "name": "INPUT__2",
+                "dataType": "TYPE_FP16",
+                "dims": [32]
+            }],
+            "max_batch_size": 8
+        }
+
+        expected_model_config1 = {
+            'instanceGroup': [{
+                'count': 3,
+                'kind': 'KIND_GPU',
+            }],
+            'maxBatchSize': 2,
+            'sequenceBatching': {},
+            'name': 'fake_model_name1',
+            'input': [{
+                "name": "INPUT__0",
+                "dataType": "TYPE_FP32",
+                "dims": ['16']
+            }]
+        }
+
+        expected_model_config2 = {
+            'dynamicBatching': {},
+            'instanceGroup': [{
+                'count': 6,
+                'kind': 'KIND_GPU',
+            }],
+            'maxBatchSize': 16,
+            'name': 'fake_model_name2',
+            'input': [{
+                "name": "INPUT__2",
+                "dataType": "TYPE_FP16",
+                "dims": ['32']
+            }]
+        }
+        # yapf: enable
+
+        mock_models = []
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            return_value=fake_base_config1,
+        ):
+            mock_models.append(
+                ModelProfileSpec(
+                    ConfigModelProfileSpec(
+                        model_name="fake_model_name1",
+                        perf_analyzer_flags={"model-version": 2},
+                    ),
+                    MagicMock(),
+                    MagicMock(),
+                    MagicMock(),
+                )
+            )
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            return_value=fake_base_config2,
+        ):
+            mock_models.append(
+                ModelProfileSpec(
+                    ConfigModelProfileSpec(
+                        model_name="fake_model_name2",
+                        perf_analyzer_flags={"model-version": 3},
+                    ),
+                    MagicMock(),
+                    MagicMock(),
+                    MagicMock(),
+                )
+            )
+
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension(
+                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+                SearchDimension(
+                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
+                ),
+            ],
+        )
+        dims.add_dimensions(
+            1,
+            [
+                SearchDimension(
+                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+                SearchDimension(
+                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
+                ),
+            ],
+        )
+
+        sc = SearchConfig(dimensions=dims, radius=5, min_initialized=2)
+        config = self._create_config()
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            mock_models,
+            {},
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        qrcg._coordinate_to_measure = Coordinate([1, 2, 4, 5])
+
+        rc = qrcg._get_next_run_config()
+
+        self.assertEqual(len(rc.model_run_configs()), 2)
+        mc1 = rc.model_run_configs()[0].model_config()
+        pc1 = rc.model_run_configs()[0].perf_config()
+        mc2 = rc.model_run_configs()[1].model_config()
+        pc2 = rc.model_run_configs()[1].perf_config()
+
+        self.assertEqual(mc1.to_dict(), expected_model_config1)
+        self.assertEqual(mc2.to_dict(), expected_model_config2)
+        self.assertEqual(pc1["concurrency-range"], 12)
+        self.assertEqual(pc1["batch-size"], 1)
+        self.assertEqual(pc1["model-version"], 2)
+        self.assertEqual(pc2["concurrency-range"], 192)
+        self.assertEqual(pc2["batch-size"], 1)
+        self.assertEqual(pc2["model-version"], 3)
+
+    def test_default_config_generation(self):
+        """
+        Test that the default config is generated correctly
+        """
+
+        fake_config = {
+            "name": "my-model",
+            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+            "max_batch_size": 4,
+        }
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "/tmp",
+            "--config-file",
+            "/tmp/my_config.yml",
+        ]
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model:
+                    perf_analyzer_flags:
+                        percentile: 96
+            """)
+        # yapf: enable
+
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            return_value=fake_config,
+        ):
+            models = [
+                ModelProfileSpec(
+                    spec=config.profile_models[0],
+                    config=config,
+                    client=MagicMock(),
+                    gpus=MagicMock(),
+                )
+            ]
+
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension(
+                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+                SearchDimension(
+                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
+                ),
+                SearchDimension(
+                    "concurrency", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+            ],
+        )
+
+        sc = SearchConfig(dimensions=dims, radius=5, min_initialized=2)
+        qrcg = QuickRunConfigGenerator(
+            sc, config, ["GPU0"], models, {}, MagicMock(), ModelVariantNameManager()
+        )
+
+        default_run_config = qrcg._create_default_run_config()
+
+        self.assertIn("--percentile=96", default_run_config.representation())
+        self.assertIn(
+            "--concurrency-range=8",
+            default_run_config.model_run_configs()[0].perf_config().representation(),
+        )
+
+    def test_default_ensemble_config_generation(self):
+        """
+        Test that the default ensemble config is generated correctly
+        """
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "/tmp",
+            "--config-file",
+            "/tmp/my_config.yml",
+        ]
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model:
+                    perf_analyzer_flags:
+                        percentile: 96
+            """)
+        # yapf: enable
+
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            side_effect=mock_ensemble_configs,
+        ):
+            models = [
+                ModelProfileSpec(
+                    spec=config.profile_models[0],
+                    config=config,
+                    client=MagicMock(),
+                    gpus=MagicMock(),
+                )
+            ]
+
+        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            side_effect=mock_ensemble_configs,
+        ):
+            ensemble_composing_models = (
+                RunConfigGeneratorFactory._create_composing_models(
+                    models, config, MagicMock(), MagicMock()
+                )
+            )
+
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            models,
+            ensemble_composing_models,
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        default_run_config = qrcg._create_default_run_config()
+        ensemble_composing_config_variants = default_run_config.model_run_configs()[
+            0
+        ].composing_config_variants()
+
+        self.assertIn("my-model", default_run_config.representation())
+        self.assertEqual(
+            ensemble_composing_config_variants[0].variant_name,
+            "fake_model_A_config_default",
+        )
+        self.assertEqual(
+            ensemble_composing_config_variants[1].variant_name,
+            "fake_model_B_config_default",
+        )
+
+    def test_default_bls_config_generation(self):
+        """
+        Test that the default BLS config is generated correctly
+        """
+
+        fake_config = {
+            "name": "my-model",
+            "platform": "pytorch",
+            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
+            "max_batch_size": 4,
+        }
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "/tmp",
+            "--config-file",
+            "/tmp/my_config.yml",
+            "--bls-composing-models",
+            "bls_composing_modelA,bls_composing_modelB",
+        ]
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model:
+                    perf_analyzer_flags:
+                        percentile: 96
+            """)
+        # yapf: enable
+
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            return_value=fake_config,
+        ):
+            models = [
+                ModelProfileSpec(
+                    spec=config.profile_models[0],
+                    config=config,
+                    client=MagicMock(),
+                    gpus=MagicMock(),
+                )
+            ]
+
+        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            return_value=fake_config,
+        ):
+            bls_composing_models = RunConfigGeneratorFactory._create_composing_models(
+                models, config, MagicMock(), MagicMock()
+            )
+
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            models,
+            bls_composing_models,
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        default_run_config = qrcg._create_default_run_config()
+        bls_composing_config_variants = default_run_config.model_run_configs()[
+            0
+        ].composing_config_variants()
+
+        self.assertIn("my-model", default_run_config.representation())
+        self.assertEqual(
+            bls_composing_config_variants[0].variant_name,
+            "bls_composing_modelA_config_default",
+        )
+        self.assertEqual(
+            bls_composing_config_variants[1].variant_name,
+            "bls_composing_modelB_config_default",
+        )
+
+    def test_get_next_run_config_ensemble(self):
+        """
+        Test that get_next_run_config() creates a proper RunConfig for ensemble
+        """
+        self._get_next_run_config_ensemble()
+
+    def test_get_next_run_config_ensemble_with_max_concurrency(self):
+        """
+        Test that get_next_run_config() creates a proper RunConfig for ensemble with a max concurrency
+        """
+        self._get_next_run_config_ensemble(max_concurrency=8)
+
+    def test_get_next_run_config_ensemble_with_min_concurrency(self):
+        """
+        Test that get_next_run_config() creates a proper RunConfig for ensemble with a min concurrency
+        """
+        self._get_next_run_config_ensemble(min_concurrency=16)
+
+    def test_get_next_run_config_bls(self):
+        """
+        Test that get_next_run_config() creates a proper RunConfig for BLS
+        """
+        self._get_next_run_config_bls()
+
+    def test_get_next_run_config_bls_with_max_concurrency(self):
+        """
+        Test that get_next_run_config() creates a proper RunConfig for BLS with a max concurrency
+        """
+        self._get_next_run_config_bls(max_concurrency=8)
+
+    def test_get_next_run_config_bls_with_min_concurrency(self):
+        """
+        Test that get_next_run_config() creates a proper RunConfig for BLS with a min concurrency
+        """
+        self._get_next_run_config_bls(min_concurrency=16)
+
+    def test_get_next_run_config_max_batch_size(self):
+        """
+        Test that run-config-search-max-model-batch-size is enforced
+
+        Sets up a case where the coordinate is [5,7], which corresponds to
+          - max_batch_size = 32 (will be capped at 16)
+          - instance_count = 8
+          - concurrency = 32*8*2 = 512 (will now be 16*8*2 = 256)
+
+        Also
+        - dynamic batching should be on
+        - existing values from the base model config should persist if they aren't overwritten
+        """
+        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
+        config = self._create_config(
+            additional_args=["--run-config-search-max-model-batch-size", "16"]
+        )
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            self._mock_models,
+            {},
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        qrcg._coordinate_to_measure = Coordinate([5, 7])
+
+        expected_model_config = {
+            "dynamicBatching": {},
+            "instanceGroup": [
+                {
+                    "count": 8,
+                    "kind": "KIND_GPU",
+                }
+            ],
+            "maxBatchSize": 16,
+            "name": "fake_model_name",
+            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
+        }
+        # yapf: enable
+
+        rc = qrcg._get_next_run_config()
+
+        self.assertEqual(len(rc.model_run_configs()), 1)
+        model_config = rc.model_run_configs()[0].model_config()
+        perf_config = rc.model_run_configs()[0].perf_config()
+
+        self.assertEqual(model_config.to_dict(), expected_model_config)
+        self.assertEqual(perf_config["concurrency-range"], 256)
+        self.assertEqual(perf_config["batch-size"], 1)
+
+    def test_get_next_run_config_max_instance_count(self):
+        """
+        Test that run-config-search-max-instance-count is enforced
+
+        Sets up a case where the coordinate is [5,7], which corresponds to
+          - max_batch_size = 32
+          - instance_count = 8 (will be capped at 4)
+          - concurrency = 32*8*2 = 512 (will now be 32*4*2 = 256)
+
+        Also
+        - dynamic batching should be on
+        - existing values from the base model config should persist if they aren't overwritten
+        """
+        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
+        config = self._create_config(
+            additional_args=["--run-config-search-max-instance-count", "4"]
+        )
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            self._mock_models,
+            {},
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        qrcg._coordinate_to_measure = Coordinate([5, 7])
+
+        expected_model_config = {
+            "dynamicBatching": {},
+            "instanceGroup": [
+                {
+                    "count": 4,
+                    "kind": "KIND_GPU",
+                }
+            ],
+            "maxBatchSize": 32,
+            "name": "fake_model_name",
+            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
+        }
+        # yapf: enable
+
+        rc = qrcg._get_next_run_config()
+
+        self.assertEqual(len(rc.model_run_configs()), 1)
+        model_config = rc.model_run_configs()[0].model_config()
+        perf_config = rc.model_run_configs()[0].perf_config()
+
+        self.assertEqual(model_config.to_dict(), expected_model_config)
+        self.assertEqual(perf_config["concurrency-range"], 256)
+        self.assertEqual(perf_config["batch-size"], 1)
+
+    def test_get_next_run_config_min_batch_size(self):
+        """
+        Test that run-config-search-min-model-batch-size is enforced
+
+        Sets up a case where the coordinate is [5,7], which corresponds to
+          - max_batch_size = 32 (will be min of 64)
+          - instance_count = 8
+          - concurrency = 32*8*2 = 512 (will now be 64*8*2 = 1024)
+
+        Also
+        - dynamic batching should be on
+        - existing values from the base model config should persist if they aren't overwritten
+        """
+        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
+        config = self._create_config(
+            additional_args=["--run-config-search-min-model-batch-size", "64"]
+        )
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            self._mock_models,
+            {},
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        qrcg._coordinate_to_measure = Coordinate([5, 7])
+
+        expected_model_config = {
+            "dynamicBatching": {},
+            "instanceGroup": [
+                {
+                    "count": 8,
+                    "kind": "KIND_GPU",
+                }
+            ],
+            "maxBatchSize": 64,
+            "name": "fake_model_name",
+            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
+        }
+        # yapf: enable
+
+        rc = qrcg._get_next_run_config()
+
+        self.assertEqual(len(rc.model_run_configs()), 1)
+        model_config = rc.model_run_configs()[0].model_config()
+        perf_config = rc.model_run_configs()[0].perf_config()
+
+        self.assertEqual(model_config.to_dict(), expected_model_config)
+        self.assertEqual(perf_config["concurrency-range"], 1024)
+        self.assertEqual(perf_config["batch-size"], 1)
+
+    def test_get_next_run_config_min_instance_count(self):
+        """
+        Test that run-config-search-min-instance-count is enforced
+
+        Sets up a case where the coordinate is [5,7], which corresponds to
+          - max_batch_size = 32
+          - instance_count = 8 (will be min of 16)
+          - concurrency = 32*8*2 = 512 (will now be 32*16*2 = 1024)
+
+        Also
+        - dynamic batching should be on
+        - existing values from the base model config should persist if they aren't overwritten
+        """
+        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
+        config = self._create_config(
+            additional_args=["--run-config-search-min-instance-count", "16"]
+        )
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            self._mock_models,
+            {},
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        qrcg._coordinate_to_measure = Coordinate([5, 7])
+
+        expected_model_config = {
+            "dynamicBatching": {},
+            "instanceGroup": [
+                {
+                    "count": 16,
+                    "kind": "KIND_GPU",
+                }
+            ],
+            "maxBatchSize": 32,
+            "name": "fake_model_name",
+            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
+        }
+        # yapf: enable
+
+        rc = qrcg._get_next_run_config()
+
+        self.assertEqual(len(rc.model_run_configs()), 1)
+        model_config = rc.model_run_configs()[0].model_config()
+        perf_config = rc.model_run_configs()[0].perf_config()
+
+        self.assertEqual(model_config.to_dict(), expected_model_config)
+        self.assertEqual(perf_config["concurrency-range"], 1024)
+        self.assertEqual(perf_config["batch-size"], 1)
+
+    def test_ensemble_in_composing_models(self):
+        """
+        Test that an ensemble as a composing model raises an exception
+        """
+        additional_args = []
+        config = self._create_config(additional_args)
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            side_effect=mock_ensemble_configs,
+        ):
+            models = [
+                ModelProfileSpec(
+                    spec=config.profile_models[0],
+                    config=config,
+                    client=MagicMock(),
+                    gpus=MagicMock(),
+                )
+            ]
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            side_effect=mock_composing_ensemble_configs,
+        ) and self.assertRaises(TritonModelAnalyzerException):
+            RunConfigGeneratorFactory._create_composing_models(
+                models, config, MagicMock(), MagicMock()
+            )
+
+    def _get_next_run_config_ensemble(self, max_concurrency=0, min_concurrency=0):
+        """
+        Test that get_next_run_config() creates a proper RunConfig for ensemble
+
+        Sets up a case where the coordinate is [1,2,4,5], which corresponds to
+          - composing model A max_batch_size = 2
+          - composing model A instance_count = 3
+          - composing model A concurrency = 2*3*2 = 12
+          - composing model B max_batch_size = 16
+          - composing model B instance_count = 6
+          - composing model B concurrency = 16*6*2 = 192
+          - ensemble model concurrency = 12 (minimum value of [12, 192])
+
+        Also,
+        - sequence batching should be on for model A
+        - dynamic batching should be on for model B
+        - existing values from the base model config should persist if they aren't overwritten
+        - existing values for perf-analyzer config should persist if they aren't overwritten
+        """
+
+        additional_args = ["--cpu-only-composing-models", "fake_model_B"]
+        if max_concurrency:
+            additional_args.append("--run-config-search-max-concurrency")
+            additional_args.append(f"{max_concurrency}")
+        if min_concurrency:
+            additional_args.append("--run-config-search-min-concurrency")
+            additional_args.append(f"{min_concurrency}")
+
+        # yapf: disable
+        expected_model_A_config_0 = {
+            'instanceGroup': [{
+                'count': 3,
+                'kind': 'KIND_GPU',
+            }],
+            'maxBatchSize': 2,
+            'sequenceBatching': {},
+            'name': 'fake_model_A',
+            'input': [{
+                "name": "INPUT__0",
+                "dataType": "TYPE_FP32",
+                "dims": ['16']
+            }]
+        }
+
+        expected_model_B_config_0 = {
+            'dynamicBatching': {},
+            'instanceGroup': [{
+                'count': 6,
+                'kind': 'KIND_CPU',
+            }],
+            'maxBatchSize': 16,
+            'name': 'fake_model_B',
+            'input': [{
+                "name": "INPUT__2",
+                "dataType": "TYPE_FP16",
+                "dims": ['32']
+            }]
+        }
+        # yapf: enable
+
+        config = self._create_config(additional_args)
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            side_effect=mock_ensemble_configs,
+        ):
+            models = [
+                ModelProfileSpec(
+                    spec=config.profile_models[0],
+                    config=config,
+                    client=MagicMock(),
+                    gpus=MagicMock(),
+                )
+            ]
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            side_effect=mock_ensemble_configs,
+        ):
+            ensemble_composing_models = (
+                RunConfigGeneratorFactory._create_composing_models(
+                    models, config, MagicMock(), MagicMock()
+                )
+            )
+
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension(
+                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+                SearchDimension(
+                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
+                ),
+            ],
+        )
+        dims.add_dimensions(
+            1,
+            [
+                SearchDimension(
+                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+                SearchDimension(
+                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
+                ),
+            ],
+        )
+
+        sc = SearchConfig(dimensions=dims, radius=5, min_initialized=2)
+
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            models,
+            ensemble_composing_models,
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        qrcg._coordinate_to_measure = Coordinate([1, 2, 4, 5])
+
+        run_config = qrcg._get_next_run_config()
+
+        self.assertEqual(len(run_config.model_run_configs()), 1)
+        self.assertEqual(
+            len(run_config.model_run_configs()[0].composing_config_variants()), 2
+        )
+
+        model_config = run_config.model_run_configs()[0].model_config()
+        perf_config = run_config.model_run_configs()[0].perf_config()
+        composing_model_A_config_0 = (
+            run_config.model_run_configs()[0]
+            .composing_config_variants()[0]
+            .model_config
+        )
+        composing_model_B_config_0 = (
+            run_config.model_run_configs()[0]
+            .composing_config_variants()[1]
+            .model_config
+        )
+
+        self.assertIsInstance(model_config, ModelConfig)
+        self.assertEqual(
+            composing_model_A_config_0.to_dict(), expected_model_A_config_0
+        )
+        self.assertEqual(
+            composing_model_B_config_0.to_dict(), expected_model_B_config_0
+        )
+
+        if max_concurrency:
+            self.assertEqual(perf_config["concurrency-range"], max_concurrency)
+        elif min_concurrency:
+            self.assertEqual(perf_config["concurrency-range"], min_concurrency)
+        else:
+            self.assertEqual(perf_config["concurrency-range"], 12)
+
+        self.assertEqual(perf_config["batch-size"], 1)
+
+    def _get_next_run_config_bls(self, max_concurrency=0, min_concurrency=0):
+        """
+        Test that get_next_run_config() creates a proper RunConfig for BLS
+
+        Sets up a case where the coordinate is [1,2,3,4,5,6], which corresponds to
+          - model max_batch_size = 2
+          - model instance count = 3
+          - composing model 1 max_batch_size = 8
+          - composing model 1 instance_count = 5
+          - composing model 2 max_batch_size = 32
+          - composing model 2 instance_count = 7
+
+        Concurrency calculations are as follows:
+          - model concurrency = 2*3*2 = 12
+          - composing model 1 concurrency = 8*5*2 = 80
+          - composing model 2 concurrency = 32*7*2 = 448
+          - model concurrency = 12 (minimum value of [12, 80, 448])
+
+        Also,
+        - sequence batching should be on for model 1
+        - dynamic batching should be on for model 2
+        - existing values from the base model config should persist if they aren't overwritten
+        - existing values for perf-analyzer config should persist if they aren't overwritten
+        """
+
+        additional_args = ["--bls-composing-models", "fake_model_A,fake_model_B"]
+        if max_concurrency:
+            additional_args.append("--run-config-search-max-concurrency")
+            additional_args.append(f"{max_concurrency}")
+        if min_concurrency:
+            additional_args.append("--run-config-search-min-concurrency")
+            additional_args.append(f"{min_concurrency}")
+
+        # yapf: disable
+        expected_model_config = {
+            'instanceGroup': [{
+                'count': 3,
+                'kind': 'KIND_GPU',
+            }],
+            'maxBatchSize': 2,
+            'dynamicBatching': {},
+            'name': 'my-model',
+            'platform': 'bls',
+            'input': [{
+                "name": "INPUT__0",
+                "dataType": "TYPE_FP32",
+                "dims": ['16']
+            }]
+        }
+
+        expected_composing_model_config0 = {
+            'instanceGroup': [{
+                'count': 5,
+                'kind': 'KIND_GPU',
+            }],
+            'maxBatchSize': 8,
+            'sequenceBatching': {},
+            'name': 'fake_model_A',
+            'input': [{
+                "name": "INPUT__0",
+                "dataType": "TYPE_FP32",
+                "dims": ['16']
+            }]
+        }
+
+        expected_composing_model_config1 = {
+            'dynamicBatching': {},
+            'instanceGroup': [{
+                'count': 7,
+                'kind': 'KIND_GPU',
+            }],
+            'maxBatchSize': 32,
+            'name': 'fake_model_B',
+            'input': [{
+                "name": "INPUT__2",
+                "dataType": "TYPE_FP16",
+                "dims": ['32']
+            }]
+        }
+        # yapf: enable
+
+        config = self._create_config(additional_args)
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            side_effect=mock_bls_configs,
+        ):
+            models = [
+                ModelProfileSpec(
+                    spec=config.profile_models[0],
+                    config=config,
+                    client=MagicMock(),
+                    gpus=MagicMock(),
+                )
+            ]
+
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension(
+                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+                SearchDimension(
+                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
+                ),
+            ],
+        )
+        dims.add_dimensions(
+            1,
+            [
+                SearchDimension(
+                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+                SearchDimension(
+                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
+                ),
+            ],
+        )
+        dims.add_dimensions(
+            2,
+            [
+                SearchDimension(
+                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
+                ),
+                SearchDimension(
+                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
+                ),
+            ],
+        )
+
+        sc = SearchConfig(dimensions=dims, radius=5, min_initialized=2)
+
+        with patch(
+            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
+            side_effect=mock_bls_configs,
+        ):
+            bls_composing_models = RunConfigGeneratorFactory._create_composing_models(
+                models, config, MagicMock(), MagicMock()
+            )
+
+        qrcg = QuickRunConfigGenerator(
+            sc,
+            config,
+            MagicMock(),
+            models,
+            bls_composing_models,
+            MagicMock(),
+            ModelVariantNameManager(),
+        )
+
+        qrcg._coordinate_to_measure = Coordinate([1, 2, 3, 4, 5, 6])
+
+        run_config = qrcg._get_next_run_config()
+
+        self.assertEqual(len(run_config.model_run_configs()), 1)
+
+        self.assertEqual(
+            len(run_config.model_run_configs()[0].composing_config_variants()), 2
+        )
+
+        model_config = run_config.model_run_configs()[0].model_config()
+        perf_config = run_config.model_run_configs()[0].perf_config()
+        composing_model_config0 = (
+            run_config.model_run_configs()[0]
+            .composing_config_variants()[0]
+            .model_config
+        )
+        composing_model_config1 = (
+            run_config.model_run_configs()[0]
+            .composing_config_variants()[1]
+            .model_config
+        )
+
+        self.assertEqual(model_config.to_dict(), expected_model_config)
+        self.assertEqual(
+            composing_model_config0.to_dict(), expected_composing_model_config0
+        )
+        self.assertEqual(
+            composing_model_config1.to_dict(), expected_composing_model_config1
+        )
+
+        if max_concurrency:
+            self.assertEqual(perf_config["concurrency-range"], max_concurrency)
+        elif min_concurrency:
+            self.assertEqual(perf_config["concurrency-range"], min_concurrency)
+        else:
+            self.assertEqual(perf_config["concurrency-range"], 12)
+
+        self.assertEqual(perf_config["batch-size"], 1)
+
+    def _create_config(self, additional_args=[]):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "/tmp",
+            "--config-file",
+            "/tmp/my_config.yml",
+        ]
+
+        for arg in additional_args:
+            args.append(arg)
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        return config
+
+    def tearDown(self):
+        patch.stopall()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_record_aggregator.py b/src/c++/perf_analyzer/tests/test_record_aggregator.py
new file mode 100755
index 000000000..f95165296
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_record_aggregator.py
@@ -0,0 +1,175 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.record.record_aggregator import RecordAggregator
+from model_analyzer.record.types.gpu_utilization import GPUUtilization
+from model_analyzer.record.types.perf_latency_p99 import PerfLatencyP99
+from model_analyzer.record.types.perf_throughput import PerfThroughput
+
+from .common import test_result_collector as trc
+
+
+class TestRecordAggregatorMethods(trc.TestResultCollector):
+    def tearDown(self):
+        patch.stopall()
+
+    def test_insert(self):
+        record_aggregator = RecordAggregator()
+
+        self.assertEqual(record_aggregator.total(), 0)
+
+        throughput_record = PerfThroughput(5)
+        record_aggregator.insert(throughput_record)
+
+        # Assert record is added
+        self.assertEqual(record_aggregator.total(), 1)
+
+    def test_record_types(self):
+        record_aggregator = RecordAggregator()
+
+        throughput_record = PerfThroughput(5)
+        record_aggregator.insert(throughput_record)
+
+        self.assertEqual(record_aggregator.record_types()[0], PerfThroughput)
+
+    def test_filter_records_default(self):
+        record_aggregator = RecordAggregator()
+
+        # insert throughput record and check its presence
+        throughput_record = PerfThroughput(5)
+        record_aggregator.insert(throughput_record)
+
+        # Get the record
+        retrieved_records = record_aggregator.filter_records().get_records()
+        retrieved_throughput = retrieved_records[PerfThroughput][0]
+
+        self.assertIsInstance(
+            retrieved_throughput,
+            PerfThroughput,
+            msg="Record types do not match after filter_records",
+        )
+
+        self.assertEqual(
+            retrieved_throughput.value(),
+            throughput_record.value(),
+            msg="Values do not match after filter_records",
+        )
+
+    def test_filter_records_filtered(self):
+        record_aggregator = RecordAggregator()
+
+        # Test for malformed inputs
+        with self.assertRaises(Exception):
+            record_aggregator.filter_records(filters=[(lambda x: False)])
+        with self.assertRaises(Exception):
+            record_aggregator.filter_records(
+                record_types=[None, None], filters=[(lambda x: False)]
+            )
+
+        # Insert 3 throughputs
+        record_aggregator.insert(PerfThroughput(5))
+        record_aggregator.insert(PerfThroughput(1))
+        record_aggregator.insert(PerfThroughput(10))
+
+        # Test get with filters
+        retrieved_records = record_aggregator.filter_records(
+            record_types=[PerfThroughput], filters=[(lambda v: v.value() >= 5)]
+        ).get_records()
+
+        # Should return 2 records
+        self.assertEqual(len(retrieved_records[PerfThroughput]), 2)
+        retrieved_values = [
+            record.value() for record in retrieved_records[PerfThroughput]
+        ]
+        self.assertIn(5, retrieved_values)
+        self.assertIn(10, retrieved_values)
+
+        # Insert 2 Latency records
+        record_aggregator.insert(PerfLatencyP99(3))
+        record_aggregator.insert(PerfLatencyP99(6))
+
+        # Test get with multiple headers
+        retrieved_records = record_aggregator.filter_records(
+            record_types=[PerfLatencyP99, PerfThroughput],
+            filters=[(lambda v: v.value() == 3), (lambda v: v.value() < 5)],
+        ).get_records()
+
+        retrieved_values = {
+            record_type: [record.value() for record in retrieved_records[record_type]]
+            for record_type in [PerfLatencyP99, PerfThroughput]
+        }
+
+        self.assertEqual(len(retrieved_records[PerfLatencyP99]), 1)
+        self.assertIn(3, retrieved_values[PerfLatencyP99])
+
+        self.assertEqual(len(retrieved_records[PerfThroughput]), 1)
+        self.assertIn(1, retrieved_values[PerfThroughput])
+
+    def test_groupby(self):
+        record_aggregator = RecordAggregator()
+        # Insert 3 throughputs
+        record_aggregator.insert(PerfThroughput(5, timestamp=0))
+        record_aggregator.insert(PerfThroughput(1, timestamp=1))
+        record_aggregator.insert(PerfThroughput(10, timestamp=1))
+
+        def groupby_criteria(record):
+            return record.timestamp()
+
+        records = record_aggregator.groupby([PerfThroughput], groupby_criteria)
+        self.assertEqual(list(records[PerfThroughput]), [0, 1])
+        self.assertEqual(
+            list(records[PerfThroughput].values()),
+            [PerfThroughput(5.0), PerfThroughput(10.0)],
+        )
+
+        records = record_aggregator.groupby([PerfThroughput], groupby_criteria)
+        self.assertEqual(list(records[PerfThroughput]), [0, 1])
+        self.assertEqual(
+            list(records[PerfThroughput].values()),
+            [PerfThroughput(5.0), PerfThroughput(10.0)],
+        )
+
+    def test_aggregate(self):
+        record_aggregator = RecordAggregator()
+
+        # Insert 10 records
+        for i in range(10):
+            record_aggregator.insert(PerfThroughput(i))
+
+        for i in range(10):
+            record_aggregator.insert(GPUUtilization(i))
+        # Aggregate them with max, min and average
+        max_vals = record_aggregator.aggregate(record_types=[PerfThroughput])
+        avg_vals = record_aggregator.aggregate(record_types=[GPUUtilization])
+
+        self.assertEqual(
+            max_vals[PerfThroughput],
+            PerfThroughput(9),
+            msg="Aggregation failed with max",
+        )
+
+        self.assertEqual(
+            avg_vals[GPUUtilization],
+            GPUUtilization(4.5),
+            msg="Aggregation failed with max",
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_record_types.py b/src/c++/perf_analyzer/tests/test_record_types.py
new file mode 100755
index 000000000..4bd6d8b32
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_record_types.py
@@ -0,0 +1,191 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.record.record import RecordType
+
+from .common import test_result_collector as trc
+
+
+class TestRecordAggregatorMethods(trc.TestResultCollector):
+    """
+    The record types in the model-analyzer.record.types package are contextual
+        when it uses 'less than' (<) and 'greater than' (>) operators.
+
+    The 'less than' and 'greater than' operators are overloaded to
+        mean 'worse than' and 'better than' respectively.
+
+    Some record types treat MORE as better
+        (eg, gpu_free_memory, cpu_available_ram)
+    Other record types treat LESS as better
+        (eg, gpu_used_memory, cpu_used_ram)
+
+    So, when comparing two objects of type 'cpu_used_ram'
+        12 > 13 is actually true, since 12 'is better than' 13.
+    """
+
+    def setUp(self):
+        record_types = RecordType.get_all_record_types()
+        self.all_record_types = record_types.values()
+        self.less_is_better_types = {
+            record_types[k]
+            for k in [
+                "perf_latency_avg",
+                "perf_latency_p90",
+                "perf_latency_p95",
+                "perf_latency_p99",
+                "gpu_used_memory",
+                "cpu_used_ram",
+                "perf_server_compute_infer",
+                "perf_latency",
+                "perf_server_queue",
+                "perf_client_response_wait",
+                "perf_server_compute_output",
+                "perf_client_send_recv",
+                "perf_server_compute_input",
+                "gpu_power_usage",
+            ]
+        }
+        self.more_is_better_types = {
+            record_types[k]
+            for k in [
+                "perf_throughput",
+                "gpu_free_memory",
+                "gpu_utilization",
+                "cpu_available_ram",
+                "gpu_total_memory",
+            ]
+        }
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_counts(self):
+        """
+        Make sure that all 'worse than' and 'better than' tests are tested
+        """
+        total_count = len(self.all_record_types)
+        less_is_better_count = len(self.less_is_better_types)
+        more_is_better_count = len(self.more_is_better_types)
+        self.assertEqual(total_count, less_is_better_count + more_is_better_count)
+
+    def test_add(self):
+        """
+        Test __add__ function for
+        each record type
+        """
+
+        for record_type in self.all_record_types:
+            metric1 = record_type(value=5)
+            metric2 = record_type(value=9)
+            metric3 = metric1 + metric2
+            self.assertIsInstance(metric3, record_type)
+            self.assertEqual(metric3.value(), 14)
+
+    def test_sub(self):
+        """
+        Test __sub__ function for
+        each record type
+        """
+
+        for record_type in self.all_record_types:
+            metric1 = record_type(value=10)
+            metric2 = record_type(value=3)
+            metric3 = metric1 - metric2
+            self.assertIsInstance(metric3, record_type)
+            if record_type in self.less_is_better_types:
+                self.assertEqual(metric3.value(), -7)
+            elif record_type in self.more_is_better_types:
+                self.assertEqual(metric3.value(), 7)
+
+    def test_mult(self):
+        """
+        Test __mult__ function for
+        each record type
+        """
+
+        for record_type in self.all_record_types:
+            metric1 = record_type(value=6)
+            metric2 = metric1 * 2
+            self.assertIsInstance(metric2, record_type)
+            self.assertEqual(metric2.value(), 12)
+
+    def test_div(self):
+        """
+        Test __div__ function for
+        each record type
+        """
+
+        for record_type in self.all_record_types:
+            metric1 = record_type(value=60)
+            metric2 = metric1 / 12
+            self.assertIsInstance(metric2, record_type)
+            self.assertEqual(metric2.value(), 5)
+
+    def test_compare(self):
+        """
+        Test __lt__, __eq__, __gt__
+        functions for each record type
+        """
+
+        for record_type in self.all_record_types:
+            metric1 = record_type(value=10.6)
+            metric2 = record_type(value=3.2)
+
+            # Test __lt__ (True if 1 worse than 2)
+            if record_type in self.less_is_better_types:
+                self.assertLess(metric1, metric2)
+            elif record_type in self.more_is_better_types:
+                self.assertLess(metric2, metric1)
+
+            # Test __gt__ (True if 1 better than 2)
+            if record_type in self.less_is_better_types:
+                self.assertLess(metric1, metric2)
+            elif record_type in self.more_is_better_types:
+                self.assertGreater(metric1, metric2)
+
+            # Test __eq__
+            metric1 = record_type(value=12)
+            metric2 = record_type(value=12)
+            self.assertEqual(metric1, metric2)
+
+    def test_value_function(self):
+        avg_value = RecordType.get_all_record_types()[
+            "perf_latency_p99"
+        ].value_function()([10, 50, 100, 40])
+
+        total_value = RecordType.get_all_record_types()[
+            "perf_throughput"
+        ].value_function()([10, 50, 100, 40])
+
+        self.assertEqual(avg_value, 50)
+        self.assertEqual(total_value, 200)
+
+    def test_calculate_percentage_gain(self):
+        for record_type in self.all_record_types:
+            metric1 = record_type(value=10)
+            metric2 = record_type(value=5)
+
+            if record_type in self.less_is_better_types:
+                self.assertEqual(metric1.calculate_percentage_gain(metric2), -50)
+            else:
+                self.assertEqual(metric1.calculate_percentage_gain(metric2), 100)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_remote_monitor.py b/src/c++/perf_analyzer/tests/test_remote_monitor.py
new file mode 100755
index 000000000..6003c3f17
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_remote_monitor.py
@@ -0,0 +1,174 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import time
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.monitor.remote_monitor import RemoteMonitor
+from model_analyzer.record.types.gpu_free_memory import GPUFreeMemory
+from model_analyzer.record.types.gpu_power_usage import GPUPowerUsage
+from model_analyzer.record.types.gpu_used_memory import GPUUsedMemory
+from model_analyzer.record.types.gpu_utilization import GPUUtilization
+from tests.mocks.mock_requests import MockRequests
+
+from .common import test_result_collector as trc
+
+TEST_DEVICE_NAME = "TEST_DEVICE_NAME"
+TEST_DEVICE_ID = 0
+TEST_METRICS_URL = "localhost:8002"
+
+TEST_POWER_USAGE = 20
+TEST_GPU_MEMORY_USAGE = 400000000
+TEST_GPU_UTILIZATION = 0.5
+TEST_TOTAL_BYTES = 1000000000
+
+TEST_METRICS_RESPONSE = bytes(
+    "# HELP nv_inference_request_success Number of successful"
+    "inference requests, all batch sizes\n# TYPE nv_inference_request_success counter\n# "
+    "HELP nv_inference_request_failure Number of failed inference requests, all batch sizes\n# "
+    "TYPE nv_inference_request_failure counter\n# HELP nv_inference_count Number of inferences "
+    "performed\n# TYPE nv_inference_count counter\n# HELP nv_inference_exec_count Number of model"
+    "executions performed\n# TYPE nv_inference_exec_count counter\n# "
+    "HELP nv_inference_request_duration_us Cumulative inference request duration in microseconds\n# "
+    "TYPE nv_inference_request_duration_us counter\n# HELP nv_inference_queue_duration_us Cumulative "
+    "inference queuing duration in microseconds\n# TYPE nv_inference_queue_duration_us counter\n# HELP "
+    "nv_inference_compute_input_duration_us Cumulative compute input duration in microseconds\n# TYPE "
+    "nv_inference_compute_input_duration_us counter\n# HELP nv_inference_compute_infer_duration_us Cumulative "
+    "compute inference duration in microseconds\n# TYPE nv_inference_compute_infer_duration_us counter\n# HELP "
+    "nv_inference_compute_output_duration_us Cumulative inference compute output duration in microseconds\n# TYPE "
+    "nv_inference_compute_output_duration_us counter\n# HELP nv_gpu_utilization GPU utilization rate [0.0 - 1.0)\n# "
+    'TYPE nv_gpu_utilization gauge\nnv_gpu_utilization{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} '
+    f"{TEST_GPU_UTILIZATION}\n# HELP nv_gpu_memory_total_bytes GPU total memory, in bytes\n# TYPE nv_gpu_memory_total_bytes "
+    'gauge\nnv_gpu_memory_total_bytes{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} '
+    f"{TEST_TOTAL_BYTES}\n# HELP nv_gpu_memory_used_bytes GPU used memory, in bytes\n# TYPE "
+    'nv_gpu_memory_used_bytes gauge\nnv_gpu_memory_used_bytes{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} '
+    f"{TEST_GPU_MEMORY_USAGE}\n# HELP nv_gpu_power_usage GPU power usage in watts\n# TYPE nv_gpu_power_usage "
+    'gauge\nnv_gpu_power_usage{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} '
+    f"{TEST_POWER_USAGE}\n# HELP nv_gpu_power_limit GPU power management limit in watts\n# TYPE "
+    'nv_gpu_power_limit gauge\nnv_gpu_power_limit{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} 280.000000\n# '
+    "HELP nv_energy_consumption GPU energy consumption in joules since the Triton Server started\n# TYPE nv_energy_consumption "
+    'counter\nnv_energy_consumption{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} 1474.042000\n',
+    encoding="ascii",
+)
+
+
+class TestRemoteMonitor(trc.TestResultCollector):
+    def setUp(self):
+        self.mock_requests = MockRequests(
+            mock_paths=["model_analyzer.monitor.remote_monitor"]
+        )
+        self.mock_requests.start()
+        self.mock_requests.set_get_request_response(TEST_METRICS_RESPONSE)
+
+    def test_record_memory(self):
+        # One measurement every 0.1 seconds
+        frequency = 0.1
+        monitoring_time = 0.1
+        metrics = [GPUUsedMemory, GPUFreeMemory]
+        gpu_monitor = RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
+        gpu_monitor.start_recording_metrics()
+        time.sleep(monitoring_time)
+        records = gpu_monitor.stop_recording_metrics()
+
+        # Assert instance types
+        for record in records:
+            self.assertIsInstance(record.device_uuid(), str)
+            self.assertIsInstance(record.value(), float)
+            if isinstance(record, GPUFreeMemory):
+                self.assertEqual(
+                    record.value(), (TEST_TOTAL_BYTES - TEST_GPU_MEMORY_USAGE) // 1e6
+                )
+            else:
+                self.assertEqual(record.value(), TEST_GPU_MEMORY_USAGE // 1e6)
+            self.assertIsInstance(record.timestamp(), int)
+
+        # The number of records should be dividable by number of metrics
+        self.assertEqual(len(records) % len(metrics), 0)
+        self.assertGreater(len(records), 0)
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            gpu_monitor.stop_recording_metrics()
+
+        gpu_monitor.destroy()
+
+        metrics = ["UndefinedTag"]
+        with self.assertRaises(TritonModelAnalyzerException):
+            RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
+
+    def test_record_power(self):
+        # One measurement every 0.01 seconds
+        frequency = 0.1
+        monitoring_time = 0.1
+        metrics = [GPUPowerUsage]
+        gpu_monitor = RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
+        gpu_monitor.start_recording_metrics()
+        time.sleep(monitoring_time)
+        records = gpu_monitor.stop_recording_metrics()
+
+        # Assert instance types
+        for record in records:
+            self.assertIsInstance(record.device_uuid(), str)
+            self.assertIsInstance(record.value(), float)
+            self.assertEqual(record.value(), TEST_POWER_USAGE)
+            self.assertIsInstance(record.timestamp(), int)
+
+        # The number of records should be dividable by number of metrics
+        self.assertEqual(len(records) % len(metrics), 0)
+        self.assertGreater(len(records), 0)
+
+        gpu_monitor.destroy()
+
+    def test_record_utilization(self):
+        # One measurement every 0.01 seconds
+        frequency = 0.1
+        monitoring_time = 0.1
+        metrics = [GPUUtilization]
+        gpu_monitor = RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
+        gpu_monitor.start_recording_metrics()
+        time.sleep(monitoring_time)
+        records = gpu_monitor.stop_recording_metrics()
+
+        # Assert instance types
+        for record in records:
+            self.assertIsInstance(record.device_uuid(), str)
+            self.assertIsInstance(record.value(), float)
+            self.assertLessEqual(record.value(), 100)
+            self.assertEqual(record.value(), TEST_GPU_UTILIZATION * 100)
+            self.assertIsInstance(record.timestamp(), int)
+
+        # The number of records should be dividable by number of metrics
+        self.assertEqual(len(records) % len(metrics), 0)
+        self.assertGreater(len(records), 0)
+
+        gpu_monitor.destroy()
+
+    def test_immediate_start_stop(self):
+        frequency = 1
+        metrics = [GPUUsedMemory, GPUFreeMemory]
+        gpu_monitor = RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
+        gpu_monitor.start_recording_metrics()
+        gpu_monitor.stop_recording_metrics()
+        gpu_monitor.destroy()
+
+    def tearDown(self):
+        patch.stopall()
+        self.mock_requests.stop()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_report_manager.py b/src/c++/perf_analyzer/tests/test_report_manager.py
new file mode 100755
index 000000000..6a4f26616
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_report_manager.py
@@ -0,0 +1,642 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.run.model_run_config import ModelRunConfig
+from model_analyzer.config.run.run_config import RunConfig
+from model_analyzer.perf_analyzer.perf_config import PerfAnalyzerConfig
+from model_analyzer.reports.report_manager import ReportManager
+from model_analyzer.result.constraint_manager import ConstraintManager
+from model_analyzer.result.result_manager import ResultManager
+from model_analyzer.result.run_config_result_comparator import RunConfigResultComparator
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+from model_analyzer.triton.model.model_config import ModelConfig
+from model_analyzer.triton.model.model_config_variant import ModelConfigVariant
+
+from .common import test_result_collector as trc
+from .common.test_utils import (
+    construct_constraint_manager,
+    construct_run_config_measurement,
+    evaluate_mock_config,
+)
+from .mocks.mock_io import MockIOMethods
+from .mocks.mock_json import MockJSONMethods
+from .mocks.mock_matplotlib import MockMatplotlibMethods
+from .mocks.mock_os import MockOSMethods
+
+
+class TestReportManager(trc.TestResultCollector):
+    def _init_managers(
+        self,
+        models="test_model",
+        num_configs_per_model=10,
+        mode="online",
+        subcommand="profile",
+        report_gpu_metrics=False,
+    ):
+        args = ["model-analyzer", subcommand, "-f", "path-to-config-file"]
+        if subcommand == "profile":
+            args.extend(["--profile-models", models])
+            args.extend(["--model-repository", "/tmp"])
+        else:
+            args.extend(["--report-model-configs", models])
+
+        if report_gpu_metrics:
+            args.extend(["--always-report-gpu-metrics"])
+
+        yaml_str = (
+            """
+            num_configs_per_model: """
+            + str(num_configs_per_model)
+            + """
+            client_protocol: grpc
+            export_path: /test/export/path
+            constraints:
+              perf_latency_p99:
+                max: 100
+              gpu_used_memory:
+                max: 10000
+        """
+        )
+        config = evaluate_mock_config(args, yaml_str, subcommand=subcommand)
+        state_manager = AnalyzerStateManager(config=config, server=None)
+
+        gpu_info = {"gpu_uuid": {"name": "fake_gpu_name", "total_memory": 1024000000}}
+        constraint_manager = ConstraintManager(config=config)
+
+        self.result_manager = ResultManager(
+            config=config,
+            state_manager=state_manager,
+            constraint_manager=constraint_manager,
+        )
+        self.report_manager = ReportManager(
+            mode=mode,
+            config=config,
+            gpu_info=gpu_info,
+            result_manager=self.result_manager,
+            constraint_manager=constraint_manager,
+        )
+
+    def _add_result_measurement(
+        self,
+        model_config_name,
+        model_name,
+        avg_gpu_metrics,
+        avg_non_gpu_metrics,
+        result_comparator,
+        cpu_only=False,
+        add_to_results_only=False,
+    ):
+        config_pb = self.model_config.copy()
+        config_pb["name"] = model_config_name
+        model_config = ModelConfig.create_from_dictionary(config_pb)
+        model_config_variant = ModelConfigVariant(
+            model_config, model_config_name, cpu_only
+        )
+
+        measurement = construct_run_config_measurement(
+            model_name=model_name,
+            model_config_names=[model_config_name],
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=avg_gpu_metrics,
+            non_gpu_metric_values=[avg_non_gpu_metrics],
+            metric_objectives=result_comparator._metric_weights,
+            model_config_weights=[1],
+        )
+
+        perf_config = PerfAnalyzerConfig()
+        perf_config.update_config({"model-name": model_config_name})
+        mrc = ModelRunConfig(model_name, model_config_variant, perf_config)
+        run_config = RunConfig({})
+        run_config.add_model_run_config(mrc)
+
+        if add_to_results_only:
+            self.result_manager._add_rcm_to_results(run_config, measurement)
+        else:
+            self.result_manager.add_run_config_measurement(run_config, measurement)
+
+    def setUp(self):
+        self.model_config = {
+            "platform": "tensorflow_graphdef",
+            "instance_group": [{"count": 1, "kind": "KIND_GPU"}],
+            "max_batch_size": 8,
+            "dynamic_batching": {},
+        }
+
+        self.os_mock = MockOSMethods(
+            mock_paths=[
+                "model_analyzer.reports.report_manager",
+                "model_analyzer.plots.plot_manager",
+                "model_analyzer.state.analyzer_state_manager",
+                "model_analyzer.config.input.config_utils",
+                "model_analyzer.config.input.config_command_profile",
+                "model_analyzer.config.input.config_command_report",
+            ]
+        )
+        self.os_mock.start()
+        self.os_mock.set_os_path_isfile_return_value(False)
+        # Required patch ordering here
+        # html_report must be patched before pdf_report
+        # Likely due to patching dealing with parent + child classes
+        self.io_mock = MockIOMethods(
+            mock_paths=[
+                "model_analyzer.reports.html_report",
+                "model_analyzer.reports.pdf_report",
+                "model_analyzer.state.analyzer_state_manager",
+            ],
+            read_data=[bytes(">:(".encode("ascii"))],
+        )
+        self.io_mock.start()
+        self.matplotlib_mock = MockMatplotlibMethods()
+        self.matplotlib_mock.start()
+        self.json_mock = MockJSONMethods()
+        self.json_mock.start()
+
+    def test_add_results(self, *args):
+        for mode in ["online", "offline"]:
+            self._init_managers(
+                models="test_model1,test_model2", mode=mode, subcommand="profile"
+            )
+            result_comparator = RunConfigResultComparator(
+                metric_objectives_list=[{"perf_throughput": 10}], model_weights=[1]
+            )
+
+            avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+
+            for i in range(10):
+                avg_non_gpu_metrics = {
+                    "perf_throughput": 100 + 10 * i,
+                    "perf_latency_p99": 4000,
+                    "cpu_used_ram": 1000,
+                }
+                self._add_result_measurement(
+                    f"test_model1_config_report_{i}",
+                    "test_model1",
+                    avg_gpu_metrics,
+                    avg_non_gpu_metrics,
+                    result_comparator,
+                    add_to_results_only=False,
+                )
+
+            for i in range(5):
+                avg_non_gpu_metrics = {
+                    "perf_throughput": 200 + 10 * i,
+                    "perf_latency_p99": 4000,
+                    "cpu_used_ram": 1000,
+                }
+                self._add_result_measurement(
+                    f"test_model2_config_report_{i}",
+                    "test_model2",
+                    avg_gpu_metrics,
+                    avg_non_gpu_metrics,
+                    result_comparator,
+                    add_to_results_only=False,
+                )
+
+            self.report_manager.create_summaries()
+            self.assertEqual(
+                self.report_manager.report_keys(), ["test_model1", "test_model2"]
+            )
+
+            report1_data = self.report_manager.data("test_model1")
+            report2_data = self.report_manager.data("test_model2")
+
+            self.assertEqual(len(report1_data), 10)
+            self.assertEqual(len(report2_data), 5)
+
+    @patch(
+        "model_analyzer.reports.report_manager.ReportManager._find_default_run_config_measurement",
+        return_value=100,
+    )
+    @patch(
+        "model_analyzer.reports.report_manager.ReportManager._get_objective_gain",
+        return_value=100,
+    )
+    def test_build_summary_table(self, *args):
+        for mode in ["offline", "online"]:
+            for cpu_only in [True, False]:
+                for report_gpu_metrics in [True, False]:
+                    self.subtest_build_summary_table(mode, cpu_only, report_gpu_metrics)
+
+    def subtest_build_summary_table(self, mode, cpu_only, report_gpu_metrics):
+        self._init_managers(
+            models="test_model",
+            mode=mode,
+            subcommand="profile",
+            report_gpu_metrics=report_gpu_metrics,
+        )
+        result_comparator = RunConfigResultComparator(
+            metric_objectives_list=[{"perf_throughput": 10}], model_weights=[1]
+        )
+
+        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+
+        gpu_metrics = report_gpu_metrics or not cpu_only
+
+        for i in range(10, 0, -1):
+            avg_non_gpu_metrics = {
+                "perf_throughput": 100 + 10 * i,
+                "perf_latency_p99": 4000,
+                "cpu_used_ram": 1000,
+            }
+            self._add_result_measurement(
+                f"test_model_config_{i}",
+                "test_model",
+                avg_gpu_metrics,
+                avg_non_gpu_metrics,
+                result_comparator,
+                cpu_only=not gpu_metrics,
+            )
+
+        self.report_manager.create_summaries()
+
+        summary_table, summary_sentence = self.report_manager._build_summary_table(
+            report_key="test_model",
+            num_measurements=10,
+            num_configurations=3,
+            gpu_name="TITAN RTX",
+            report_gpu_metrics=gpu_metrics,
+        )
+
+        if mode == "online":
+            objective = "maximizing throughput"
+        else:
+            objective = "minimizing latency"
+
+        if gpu_metrics:
+            expected_summary_sentence = (
+                "In 10 measurements across 3 configurations, "
+                "<strong>test_model_config_10</strong> is <strong>100%</strong> better than the default configuration "
+                f"at {objective}, under the given constraints, on GPU(s) TITAN RTX.<UL><LI> "
+                "<strong>test_model_config_10</strong>: 1 GPU instance with a max batch size of 8 on platform tensorflow_graphdef "
+                "</LI> </UL>"
+            )
+        else:
+            expected_summary_sentence = (
+                "In 10 measurements across 3 configurations, "
+                "<strong>test_model_config_10</strong> is <strong>100%</strong> better than the default configuration "
+                f"at {objective}, under the given constraints.<UL><LI> "
+                "<strong>test_model_config_10</strong>: 1 GPU instance with a max batch size of 8 on platform tensorflow_graphdef "
+                "</LI> </UL>"
+            )
+
+        self.assertEqual(len(expected_summary_sentence), len(summary_sentence))
+        self.assertEqual(expected_summary_sentence, summary_sentence)
+
+        # Get throughput index and make sure results are sorted
+        throughput_index = summary_table.headers().index("Throughput (infer/sec)")
+        model_name_index = summary_table.headers().index("Model Config Name")
+        for i in range(9):
+            current_row = summary_table.get_row_by_index(i)
+            next_row = summary_table.get_row_by_index(i + 1)
+            self.assertEqual(current_row[model_name_index], f"test_model_config_{10-i}")
+            self.assertGreaterEqual(
+                current_row[throughput_index], next_row[throughput_index]
+            )
+
+    def test_build_detailed_info(self):
+        for cpu_only in [True, False]:
+            for report_gpu_metrics in [True, False]:
+                self._subtest_build_detailed_info(cpu_only, report_gpu_metrics)
+
+    def _subtest_build_detailed_info(self, cpu_only, report_gpu_metrics):
+        self._init_managers(models="test_model_config_10", subcommand="report")
+
+        result_comparator = RunConfigResultComparator(
+            metric_objectives_list=[{"perf_throughput": 10}], model_weights=[1]
+        )
+
+        gpu_metrics = report_gpu_metrics or not cpu_only
+
+        avg_gpu_metrics = {"gpu_uuid": {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+
+        for i in range(10, 0, -1):
+            avg_non_gpu_metrics = {
+                "perf_throughput": 100 + 10 * i,
+                "perf_latency_p99": 4000,
+                "cpu_used_ram": 1000,
+            }
+            self._add_result_measurement(
+                f"test_model_config_{i}",
+                "test_model",
+                avg_gpu_metrics,
+                avg_non_gpu_metrics,
+                result_comparator,
+                cpu_only=not gpu_metrics,
+                add_to_results_only=True,
+            )
+
+        self.report_manager._add_detailed_report_data()
+        self.report_manager._build_detailed_table("test_model_config_10")
+        sentence = self.report_manager._build_detailed_info("test_model_config_10")
+
+        if gpu_metrics:
+            expected_sentence = (
+                f"The model config <strong>test_model_config_10</strong> uses 1 GPU instance with "
+                f"a max batch size of 8 and has dynamic batching enabled. 1 measurement(s) "
+                f"were obtained for the model config on GPU(s) 1 x fake_gpu_name with total memory 1.0 GB. "
+                f"This model uses the platform tensorflow_graphdef."
+            )
+        else:
+            expected_sentence = (
+                f"The model config <strong>test_model_config_10</strong> uses 1 GPU instance with "
+                f"a max batch size of 8 and has dynamic batching enabled. 1 measurement(s) "
+                f"were obtained for the model config on CPU. "
+                f"This model uses the platform tensorflow_graphdef."
+            )
+
+        self.assertEqual(expected_sentence, sentence)
+
+    @patch("model_analyzer.plots.plot_manager.PlotManager._create_update_simple_plot")
+    @patch("model_analyzer.result.result_table.ResultTable.insert_row_by_index")
+    def test_summary_default_within_top(self, add_table_fn, add_plot_fn):
+        """
+        Test summary report generation when default is in the top n configs
+
+        Creates some results where the default config is within the top n configs,
+        and then confirms that the number of entries added to plots and tables
+        is correct
+        """
+
+        default_within_top = True
+        top_n = 3
+        self._test_summary_counts(add_table_fn, add_plot_fn, default_within_top, top_n)
+
+    @patch("model_analyzer.plots.plot_manager.PlotManager._create_update_simple_plot")
+    @patch("model_analyzer.result.result_table.ResultTable.insert_row_by_index")
+    def test_summary_default_not_within_top(self, add_table_fn, add_plot_fn):
+        """
+        Test summary report generation when default is not in the top n configs
+
+        Creates some results where the default config is not within the top n configs,
+        and then confirms that the number of entries added to plots and tables
+        is correct such that it includes the default config data
+        """
+        default_within_top = False
+        top_n = 3
+        self._test_summary_counts(add_table_fn, add_plot_fn, default_within_top, top_n)
+
+    def _test_summary_counts(
+        self, add_table_fn, add_plot_fn, default_within_top, top_n
+    ):
+        """
+        Helper function to test creating summary reports and confirming that the number
+        of entries added to plots and tables is as expected
+        """
+        num_plots_in_summary_report = 2
+        num_tables_in_summary_report = 1
+        expected_config_count = top_n + 1 if not default_within_top else top_n
+        expected_plot_count = num_plots_in_summary_report * expected_config_count
+        expected_table_count = num_tables_in_summary_report * expected_config_count
+
+        self._init_managers(
+            models="test_model1", num_configs_per_model=top_n, subcommand="profile"
+        )
+        result_comparator = RunConfigResultComparator(
+            metric_objectives_list=[{"perf_throughput": 10}], model_weights=[1]
+        )
+        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+        for i in range(10):
+            # Create a bunch of fake measurement results
+            #
+            # For the 'default_within_top' case, have the throughput
+            # be decreasing for each config (so the highest throughput is the
+            # first, default config)
+            #
+            # For the default not within top case, have the throughput be
+            # increasing for each config (so the highest throughput is the last
+            # config, and the default config is the worst)
+            #
+            p99 = 20 + i
+            throughput = 100 - 10 * i if default_within_top else 200 + 10 * i
+            avg_non_gpu_metrics = {
+                "perf_throughput": throughput,
+                "perf_latency_p99": p99,
+                "cpu_used_ram": 1000,
+            }
+            name = f"test_model1_config_{i}"
+            if not i:
+                name = f"test_model1_config_default"
+            self._add_result_measurement(
+                name,
+                "test_model1",
+                avg_gpu_metrics,
+                avg_non_gpu_metrics,
+                result_comparator,
+            )
+        self.report_manager.create_summaries()
+
+        self.assertEqual(expected_plot_count, add_plot_fn.call_count)
+        self.assertEqual(expected_table_count, add_table_fn.call_count)
+
+    def test_create_instance_group_phrase(self):
+        """Test all corner cases of _create_instance_group_phrase()"""
+        self._init_managers(models="test_model", subcommand="profile")
+
+        model_config_dict = {
+            "instance_group": [
+                {
+                    "count": 1,
+                    "kind": "KIND_GPU",
+                }
+            ]
+        }
+        self._test_instance_group_helper(model_config_dict, "1 GPU instance")
+
+        model_config_dict = {
+            "instance_group": [
+                {
+                    "count": 2,
+                    "kind": "KIND_GPU",
+                }
+            ]
+        }
+        self._test_instance_group_helper(model_config_dict, "2 GPU instances")
+
+        model_config_dict = {
+            "instance_group": [
+                {
+                    "count": 3,
+                    "kind": "KIND_GPU",
+                },
+                {
+                    "count": 2,
+                    "kind": "KIND_CPU",
+                },
+            ]
+        }
+        self._test_instance_group_helper(
+            model_config_dict, "3 GPU instances and 2 CPU instances"
+        )
+
+    def _test_instance_group_helper(self, model_config_dict, expected_output):
+        model_config = ModelConfig.create_from_dictionary(model_config_dict)
+        output = self.report_manager._create_instance_group_phrase(model_config)
+        self.assertEqual(output, expected_output)
+
+    def test_gpu_name_and_memory(self):
+        """
+        Test return value of _get_gpu_stats()
+
+        Creates a scenario where 4 GPUs are visible, but only 3 are
+        used (1 measurement uses 1 and 2, other uses 1 and 4)
+
+        The function should note that 1 and 2 are the same GPU type and combine
+        them in the name string, and it should combine the total memory of the used
+        GPUs
+        """
+        gpu_info = {
+            "gpu_uuid1": {"name": "fake_gpu_name", "total_memory": 2**30 * 2},  # 2GB
+            "gpu_uuid2": {"name": "fake_gpu_name", "total_memory": 2**30 * 2},  # 2GB
+            "gpu_uuid3": {
+                "name": "fake_gpu_name_2",
+                "total_memory": 2**30 * 4,  # 4GB
+            },
+            "gpu_uuid4": {
+                "name": "fake_gpu_name_3",
+                "total_memory": 2**30 * 8,  # 8GB
+            },
+        }
+
+        report_manager = ReportManager(
+            mode=MagicMock(),
+            config=MagicMock(),
+            gpu_info=gpu_info,
+            result_manager=MagicMock(),
+            constraint_manager=MagicMock(),
+        )
+
+        avg_gpu_metrics1 = {
+            "gpu_uuid1": {"gpu_used_memory": 6000, "gpu_utilization": 60},
+            "gpu_uuid2": {"gpu_used_memory": 6000, "gpu_utilization": 60},
+        }
+
+        avg_gpu_metrics2 = {
+            "gpu_uuid1": {"gpu_used_memory": 6000, "gpu_utilization": 60},
+            "gpu_uuid4": {"gpu_used_memory": 6000, "gpu_utilization": 60},
+        }
+
+        measurement1 = construct_run_config_measurement(
+            model_name=MagicMock(),
+            model_config_names=MagicMock(),
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=avg_gpu_metrics1,
+            constraint_manager=MagicMock(),
+            non_gpu_metric_values=MagicMock(),
+            metric_objectives=MagicMock(),
+            model_config_weights=MagicMock(),
+        )
+
+        measurement2 = construct_run_config_measurement(
+            model_name=MagicMock(),
+            model_config_names=MagicMock(),
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=avg_gpu_metrics2,
+            constraint_manager=MagicMock(),
+            non_gpu_metric_values=MagicMock(),
+            metric_objectives=MagicMock(),
+            model_config_weights=MagicMock(),
+        )
+
+        measurements = [measurement1, measurement2]
+        names, max_mem = report_manager._get_gpu_stats(measurements)
+        self.assertEqual(names, "2 x fake_gpu_name, 1 x fake_gpu_name_3")
+        self.assertEqual(max_mem, "12.0 GB")
+
+    @patch(
+        "model_analyzer.reports.report_manager.ReportManager._build_constraint_strings",
+        return_value={"modelA": "Max p99 latency: 100 ms"},
+    )
+    def test_constraint_string_single_model(self, *args):
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_latency_p99:
+                    max: 100
+            """
+        )
+
+        report_manager = ReportManager(
+            mode=MagicMock(),
+            config=MagicMock(),
+            gpu_info=MagicMock(),
+            result_manager=MagicMock(),
+            constraint_manager=constraint_manager,
+        )
+        expected_constraint_str = "Max p99 latency: 100 ms"
+        actual_constraint_str = report_manager._create_constraint_string(
+            report_key="modelA"
+        )
+
+        self.assertEqual(actual_constraint_str, expected_constraint_str)
+
+    @patch(
+        "model_analyzer.reports.report_manager.ReportManager._build_constraint_strings",
+        return_value={
+            "modelA": "Max p99 latency: 100 ms",
+            "modelB": "Max p99 latency: 200 ms",
+        },
+    )
+    def test_constraint_string_multi_model(self, *args):
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_latency_p99:
+                    max: 100
+              modelB:
+                constraints:
+                  perf_latency_p99:
+                    max: 200
+            """
+        )
+
+        report_manager = ReportManager(
+            mode=MagicMock(),
+            config=MagicMock(),
+            gpu_info=MagicMock(),
+            result_manager=MagicMock(),
+            constraint_manager=constraint_manager,
+        )
+        expected_constraint_str = "<strong>modelA</strong>: Max p99 latency: 100 ms"
+        expected_constraint_str += "<br>"
+
+        for i in range(len("Constraint targets: ")):
+            expected_constraint_str += "&ensp;"
+
+        expected_constraint_str += "<strong>modelB</strong>: Max p99 latency: 200 ms"
+
+        actual_constraint_str = report_manager._create_constraint_string(
+            report_key="modelA,modelB"
+        )
+
+        self.assertEqual(actual_constraint_str, expected_constraint_str)
+
+    def tearDown(self):
+        self.matplotlib_mock.stop()
+        self.io_mock.stop()
+        self.os_mock.stop()
+        self.json_mock.stop()
+        patch.stopall()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_report_utils.py b/src/c++/perf_analyzer/tests/test_report_utils.py
new file mode 100755
index 000000000..cb22a4650
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_report_utils.py
@@ -0,0 +1,60 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.reports.report_utils import truncate_model_config_name
+
+from .common import test_result_collector as trc
+
+
+class TestReportUtils(trc.TestResultCollector):
+    """
+    Tests the report_utils functions
+    """
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_truncate_model_config_name(self):
+        """
+        Test the behavior of the truncate_longer_model_config_name function.
+        """
+
+        # Test: Shorter model config name (below 35 characters threshold)
+        model_config_name = "ensemble_model_23"
+        result = truncate_model_config_name(model_config_name)
+        self.assertEqual(model_config_name, result)
+
+        # Test: Model config name ends with 'config_#'
+        model_config_name = "long_pytorch_platform_handler_config_10"
+        result = truncate_model_config_name(model_config_name)
+        self.assertEqual(result, "long_pytorch_platform_h...config_10")
+
+        # Test: Model config name ends with 'config_default'.
+        model_config_name = "long_pytorch_platform_handler_config_default"
+        result = truncate_model_config_name(model_config_name)
+        self.assertEqual(result, "long_pytorch_platf...config_default")
+
+        # Test: Model config name includes the "config" keyword in the model name
+        model_config_name = "long_config_pytorch_platform_handler_config_128"
+        result = truncate_model_config_name(model_config_name)
+        self.assertEqual(result, "long_config_pytorch_pl...config_128")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_result_manager.py b/src/c++/perf_analyzer/tests/test_result_manager.py
new file mode 100755
index 000000000..a16e11b31
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_result_manager.py
@@ -0,0 +1,199 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.input.config_command_report import ConfigCommandReport
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.result.constraint_manager import ConstraintManager
+from model_analyzer.result.result_manager import ResultManager
+from model_analyzer.result.sorted_results import SortedResults
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+
+from .common import test_result_collector as trc
+from .common.test_utils import (
+    load_multi_model_result_manager,
+    load_single_model_result_manager,
+)
+
+
+class TestResultManager(trc.TestResultCollector):
+    def test_server_data(self):
+        """
+        Test that add_server_data() and get_server_only_data()
+        are effectively mirrored set/get functions
+        """
+
+        state_manager = AnalyzerStateManager(config=MagicMock(), server=None)
+        result_manager = ResultManager(
+            config=ConfigCommandReport(),
+            state_manager=state_manager,
+            constraint_manager=ConstraintManager(config=MagicMock()),
+        )
+
+        server_data = {"a": 5, "b": 7}
+        result_manager.add_server_data(server_data)
+
+        self.assertEqual(server_data, result_manager.get_server_only_data())
+
+    def test_measurements(self):
+        """
+        Test add_run_config_measurement and get_model_configs_run_config_measurements
+
+        Confirm that run_config_measurements are stored per-model, and then per-variant, and
+        then in a list per-representation
+        Confirm that the measurements can be read out via get_model_configs_run_config_measurements()
+        """
+
+        state_manager = AnalyzerStateManager(config=MagicMock(), server=None)
+        result_manager = ResultManager(
+            config=ConfigCommandReport(),
+            state_manager=state_manager,
+            constraint_manager=ConstraintManager(config=MagicMock()),
+        )
+
+        fake_run_config1a = MagicMock()
+        fake_run_config1a.models_name.return_value = "Model1"
+        fake_run_config1a.model_variants_name.return_value = "Model1_config_default"
+        fake_run_config1a.representation.return_value = "1a"
+
+        fake_run_config1b = MagicMock()
+        fake_run_config1b.models_name.return_value = "Model1"
+        fake_run_config1b.model_variants_name.return_value = "Model1_config_default"
+        fake_run_config1b.representation.return_value = "1b"
+
+        fake_run_config1c = MagicMock()
+        fake_run_config1c.models_name.return_value = "Model1"
+        fake_run_config1c.model_variants_name.return_value = "Model1_config_0"
+        fake_run_config1c.representation.return_value = "1c"
+
+        fake_run_config2 = MagicMock()
+        fake_run_config2.models_name.return_value = "Model2"
+        fake_run_config2.model_variants_name.return_value = "Model2_config_default"
+        fake_run_config2.representation.return_value = "2"
+
+        rcm1a = MagicMock()
+        rcm1b = MagicMock()
+        rcm1c = MagicMock()
+        rcm2 = MagicMock()
+
+        result_manager._add_rcm_to_results(
+            run_config=fake_run_config1a, run_config_measurement=rcm1a
+        )
+        result_manager._add_rcm_to_results(
+            run_config=fake_run_config1b, run_config_measurement=rcm1b
+        )
+        result_manager._add_rcm_to_results(
+            run_config=fake_run_config1c, run_config_measurement=rcm1c
+        )
+        result_manager._add_rcm_to_results(
+            run_config=fake_run_config2, run_config_measurement=rcm2
+        )
+
+        config, measurements = result_manager.get_model_configs_run_config_measurements(
+            "Model1_config_default"
+        )
+        self.assertEqual(config, fake_run_config1a)
+        self.assertEqual(2, len(measurements))
+
+        config, measurements = result_manager.get_model_configs_run_config_measurements(
+            "Model1_config_0"
+        )
+        self.assertEqual(config, fake_run_config1c)
+        self.assertEqual(1, len(measurements))
+
+        config, measurements = result_manager.get_model_configs_run_config_measurements(
+            "Model2_config_default"
+        )
+        self.assertEqual(config, fake_run_config2)
+        self.assertEqual(1, len(measurements))
+
+    def test_get_single_model_names(self):
+        """Test get_model_names for a single-model run"""
+        result_manager, _ = load_single_model_result_manager()
+        self.assertEqual(result_manager.get_model_names(), ["add_sub"])
+
+    def test_get_multi_model_names(self):
+        """Test get_model_names for a multi-model run"""
+        result_manager, _ = load_multi_model_result_manager()
+        self.assertEqual(
+            result_manager.get_model_names(), ["resnet50_libtorch,vgg19_libtorch"]
+        )
+
+    def test_get_model_sorted_results_bad_name(self):
+        """
+        Test get_model_sorted_results will assert when called
+        with name that doesn't exist
+        """
+
+        result_manager, _ = load_multi_model_result_manager()
+        with self.assertRaises(TritonModelAnalyzerException):
+            result_manager.get_model_sorted_results("SHOULD_ERROR")
+
+    def test_get_single_model_sorted_results(self):
+        """
+        Test get_model_sorted_results returns a valid list
+        with only results for that model
+        """
+        result_manager, _ = load_single_model_result_manager()
+        result_manager._add_rcm_to_results(MagicMock(), MagicMock())
+        sorted_results = result_manager.get_model_sorted_results("add_sub")
+        self.assertTrue(isinstance(sorted_results, SortedResults))
+        self.assertEqual(5, len(sorted_results.results()))
+
+    def test_get_multi_model_sorted_results(self):
+        """Test get_model_sorted_results returns a valid list"""
+        result_manager, _ = load_multi_model_result_manager()
+        sorted_results = result_manager.get_model_sorted_results(
+            "resnet50_libtorch,vgg19_libtorch"
+        )
+        self.assertTrue(isinstance(sorted_results, SortedResults))
+        self.assertEqual(7, len(sorted_results.results()))
+
+    def test_get_across_model_sorted_results(self):
+        """
+        Test get_across_model_sorted_results returns a valid list
+        with all results across all models
+        """
+        result_manager, _ = load_single_model_result_manager()
+        self._add_a_fake_result(result_manager)
+
+        sorted_results = result_manager.get_across_model_sorted_results()
+        self.assertTrue(isinstance(sorted_results, SortedResults))
+        self.assertEqual(6, len(sorted_results.results()))
+
+    def _add_a_fake_result(self, result_manager):
+        fake_model = MagicMock()
+        fake_model.model_name.return_value = "FakeModel"
+        old_profile_models = result_manager._config.profile_models
+        old_profile_models.append(fake_model)
+        result_manager._config.profile_models = old_profile_models
+        result_manager._run_comparators["FakeModel"] = result_manager._run_comparators[
+            "add_sub"
+        ]
+
+        fake_run_config = MagicMock()
+        fake_run_config.models_name.return_value = "FakeModel"
+
+        result_manager.add_run_config_measurement(fake_run_config, MagicMock())
+
+    def tearDown(self):
+        patch.stopall()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_result_table.py b/src/c++/perf_analyzer/tests/test_result_table.py
new file mode 100755
index 000000000..b7e3ee64e
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_result_table.py
@@ -0,0 +1,121 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.result.result_table import ResultTable
+
+from .common import test_result_collector as trc
+
+TEST_HEADERS = [f"Column {i}" for i in range(10)]
+TEST_COLUMN_WIDTH = 12
+TEST_WIDTHS = [TEST_COLUMN_WIDTH for i in range(10)]
+TEST_ROWS = [[f"value {i}{j}" for j in range(10)] for i in range(10)]
+TEST_TABLE_STR = (
+    "Column 0  Column 1  Column 2  Column 3  Column 4  Column 5  Column 6  Column 7  Column 8  Column 9  \n"
+    "value 00  value 01  value 02  value 03  value 04  value 05  value 06  value 07  value 08  value 09  \n"
+    "value 10  value 11  value 12  value 13  value 14  value 15  value 16  value 17  value 18  value 19  \n"
+    "value 20  value 21  value 22  value 23  value 24  value 25  value 26  value 27  value 28  value 29  \n"
+    "value 30  value 31  value 32  value 33  value 34  value 35  value 36  value 37  value 38  value 39  \n"
+    "value 40  value 41  value 42  value 43  value 44  value 45  value 46  value 47  value 48  value 49  \n"
+    "value 50  value 51  value 52  value 53  value 54  value 55  value 56  value 57  value 58  value 59  \n"
+    "value 60  value 61  value 62  value 63  value 64  value 65  value 66  value 67  value 68  value 69  \n"
+    "value 70  value 71  value 72  value 73  value 74  value 75  value 76  value 77  value 78  value 79  \n"
+    "value 80  value 81  value 82  value 83  value 84  value 85  value 86  value 87  value 88  value 89  \n"
+    "value 90  value 91  value 92  value 93  value 94  value 95  value 96  value 97  value 98  value 99  "
+)
+TEST_CSV_STR = (
+    "Column 0,Column 1,Column 2,Column 3,Column 4,Column 5,Column 6,Column 7,Column 8,Column 9\n"
+    "value 00,value 01,value 02,value 03,value 04,value 05,value 06,value 07,value 08,value 09\n"
+    "value 10,value 11,value 12,value 13,value 14,value 15,value 16,value 17,value 18,value 19\n"
+    "value 20,value 21,value 22,value 23,value 24,value 25,value 26,value 27,value 28,value 29\n"
+    "value 30,value 31,value 32,value 33,value 34,value 35,value 36,value 37,value 38,value 39\n"
+    "value 40,value 41,value 42,value 43,value 44,value 45,value 46,value 47,value 48,value 49\n"
+    "value 50,value 51,value 52,value 53,value 54,value 55,value 56,value 57,value 58,value 59\n"
+    "value 60,value 61,value 62,value 63,value 64,value 65,value 66,value 67,value 68,value 69\n"
+    "value 70,value 71,value 72,value 73,value 74,value 75,value 76,value 77,value 78,value 79\n"
+    "value 80,value 81,value 82,value 83,value 84,value 85,value 86,value 87,value 88,value 89\n"
+    "value 90,value 91,value 92,value 93,value 94,value 95,value 96,value 97,value 98,value 99"
+)
+
+
+class TestResultTableMethods(trc.TestResultCollector):
+    def tearDown(self):
+        patch.stopall()
+
+    def test_create_headers(self):
+        table = ResultTable(headers=["Column 0"])
+        self.assertEqual(table.headers(), ["Column 0"])
+        self.assertEqual(
+            table.column_widths(), [len("Column 0") + ResultTable.column_padding]
+        )
+
+    def test_add_get_methods(self):
+        table = ResultTable(headers=["Column 0", "Column 1"])
+
+        # add/get single result
+        table.insert_row_by_index(["value 0,0", "value 0,1"])
+        self.assertEqual(table.get_row_by_index(index=0), ["value 0,0", "value 0,1"])
+        table.insert_row_by_index(["value -1,0", "value -1,1"], index=0)
+        self.assertEqual(table.get_row_by_index(index=0), ["value -1,0", "value -1,1"])
+        table.insert_row_by_index(["value 1,0", "value 1,1"], index=2)
+        self.assertEqual(table.get_row_by_index(index=2), ["value 1,0", "value 1,1"])
+
+    def test_remove_methods(self):
+        table = ResultTable(headers=TEST_HEADERS)
+        # Pick a row
+        row_idx = 4
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            table.remove_row_by_index(index=row_idx)
+
+        # Fill table
+        for row in TEST_ROWS:
+            table.insert_row_by_index(row)
+
+        self.assertEqual(
+            table.get_row_by_index(index=row_idx), [f"value 4{j}" for j in range(10)]
+        )
+
+        # remove rows and check that the next one is in its place
+        table.remove_row_by_index(index=row_idx)
+        self.assertEqual(
+            table.get_row_by_index(index=row_idx), [f"value 5{j}" for j in range(10)]
+        )
+
+        # Remove all rows and then one
+        with self.assertRaises(TritonModelAnalyzerException):
+            for i in range(10):
+                table.remove_row_by_index(index=0)
+
+    def test_to_formatted_string(self):
+        table = ResultTable(headers=TEST_HEADERS)
+        for row in TEST_ROWS:
+            table.insert_row_by_index(row)
+        self.assertEqual(table.to_formatted_string(), TEST_TABLE_STR)
+
+        table = ResultTable(headers=TEST_HEADERS)
+        for row in TEST_ROWS:
+            table.insert_row_by_index(row)
+        self.assertEqual(
+            table.to_formatted_string(separator=",", ignore_widths=True), TEST_CSV_STR
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_result_table_manager.py b/src/c++/perf_analyzer/tests/test_result_table_manager.py
new file mode 100755
index 000000000..278ceb5ea
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_result_table_manager.py
@@ -0,0 +1,344 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from filecmp import cmp
+from shutil import rmtree
+from unittest.mock import patch
+
+from google.protobuf import text_format
+from tritonclient.grpc import model_config_pb2
+
+from model_analyzer.result.constraint_manager import ConstraintManager
+from model_analyzer.result.result_manager import ResultManager
+from model_analyzer.result.result_table_manager import ResultTableManager
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+
+from .common import test_result_collector as trc
+from .common.test_utils import (
+    ROOT_DIR,
+    evaluate_mock_config,
+    load_bls_result_manager,
+    load_ensemble_result_manager,
+    load_multi_model_result_manager,
+    load_request_rate_result_manager,
+    load_single_model_result_manager,
+)
+
+
+class TestResultTableManager(trc.TestResultCollector):
+    def test_single_model_csv_against_golden(self):
+        """
+        Match the csvs against the golden versions in
+        tests/common/single-model-ckpt
+        """
+        table_manager = self._create_single_model_result_table_manager()
+
+        table_manager.create_tables()
+        table_manager.tabulate_results()
+        table_manager.export_results()
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/single-model-ckpt/results/metrics-model-gpu.csv",
+                f"{ROOT_DIR}/single-model-ckpt/golden-metrics-model-gpu.csv",
+            )
+        )
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/single-model-ckpt/results/metrics-model-inference.csv",
+                f"{ROOT_DIR}/single-model-ckpt/golden-metrics-model-inference.csv",
+            )
+        )
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/single-model-ckpt/results/metrics-server-only.csv",
+                f"{ROOT_DIR}/single-model-ckpt/golden-metrics-server-only.csv",
+            )
+        )
+
+        rmtree(f"{ROOT_DIR}/single-model-ckpt/results/")
+
+    def test_multi_model_csv_against_golden(self):
+        """
+        Match the csvs against the golden versions in
+        tests/common/multi-model-ckpt
+        """
+        table_manager = self._create_multi_model_result_table_manager()
+
+        table_manager.create_tables()
+        table_manager.tabulate_results()
+        table_manager.export_results()
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/multi-model-ckpt/results/metrics-model-gpu.csv",
+                f"{ROOT_DIR}/multi-model-ckpt/golden-metrics-model-gpu.csv",
+            )
+        )
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/multi-model-ckpt/results/metrics-model-inference.csv",
+                f"{ROOT_DIR}/multi-model-ckpt/golden-metrics-model-inference.csv",
+            )
+        )
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/multi-model-ckpt/results/metrics-server-only.csv",
+                f"{ROOT_DIR}/multi-model-ckpt/golden-metrics-server-only.csv",
+            )
+        )
+
+        rmtree(f"{ROOT_DIR}/multi-model-ckpt/results/")
+
+    def test_ensemble_csv_against_golden(self):
+        """
+        Match the csvs against the golden versions in
+        tests/common/ensemble-ckpt
+        """
+        table_manager = self._create_ensemble_result_table_manager()
+
+        table_manager.create_tables()
+        table_manager.tabulate_results()
+        table_manager.export_results()
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/ensemble-ckpt/results/metrics-model-gpu.csv",
+                f"{ROOT_DIR}/ensemble-ckpt/golden-metrics-model-gpu.csv",
+            )
+        )
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/ensemble-ckpt/results/metrics-model-inference.csv",
+                f"{ROOT_DIR}/ensemble-ckpt/golden-metrics-model-inference.csv",
+            )
+        )
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/ensemble-ckpt/results/metrics-server-only.csv",
+                f"{ROOT_DIR}/ensemble-ckpt/golden-metrics-server-only.csv",
+            )
+        )
+
+        rmtree(f"{ROOT_DIR}/ensemble-ckpt/results/")
+
+    def test_request_rate_csv_against_golden(self):
+        """
+        Match the csvs against the golden versions in
+        tests/common/request-rate-ckpt
+        """
+        table_manager = self._create_request_rate_result_table_manager()
+
+        table_manager.create_tables()
+        table_manager.tabulate_results()
+        table_manager.export_results()
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/request-rate-ckpt/results/metrics-model-gpu.csv",
+                f"{ROOT_DIR}/request-rate-ckpt/golden-metrics-model-gpu.csv",
+            )
+        )
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/request-rate-ckpt/results/metrics-model-inference.csv",
+                f"{ROOT_DIR}/request-rate-ckpt/golden-metrics-model-inference.csv",
+            )
+        )
+
+        self.assertTrue(
+            cmp(
+                f"{ROOT_DIR}/request-rate-ckpt/results/metrics-server-only.csv",
+                f"{ROOT_DIR}/request-rate-ckpt/golden-metrics-server-only.csv",
+            )
+        )
+
+        rmtree(f"{ROOT_DIR}/request-rate-ckpt/results/")
+
+    def test_create_inference_table_with_backend_parameters(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "-f",
+            "config.yml",
+            "--model-repository",
+            ".",
+        ]
+        yaml_str = """
+            profile_models: profile_models
+            inference_output_fields: model_name,batch_size,backend_parameter/parameter_1,backend_parameter/parameter_2
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+        state_manager = AnalyzerStateManager(config=config, server=None)
+        result_manager = ResultManager(
+            config=config,
+            state_manager=state_manager,
+            constraint_manager=ConstraintManager(config),
+        )
+        result_table_manager = ResultTableManager(
+            config=config, result_manager=result_manager
+        )
+
+        result_table_manager._create_inference_table()
+        self.assertEqual(
+            result_table_manager._inference_output_fields,
+            [
+                "model_name",
+                "batch_size",
+                "backend_parameter/parameter_1",
+                "backend_parameter/parameter_2",
+            ],
+        )
+
+    def test_get_common_row_items_with_backend_parameters(self):
+        """
+        This tests that a metrics model inference table row can be created with
+        backend parameters included. Each backend parameter gets its own column.
+        The column name is the backend parameter key (prepended with a prefix
+        to avoid potentially overlapping with an existing column). The column
+        value is the backend parameter value.
+
+        Here is an example table:
+
+        Models (Inference):
+        Model     Model Config Path   backend_parameter/add_sub_key_1   backend_parameter/add_sub_key_2
+        add_sub   add_sub_config_2    add_sub_value_1                   add_sub_value_2
+        add_sub   add_sub_config_0    add_sub_value_1                   add_sub_value_2
+        add_sub   add_sub_config_1    add_sub_value_1                   add_sub_value_2
+
+        Each row of the metrics model inference table corresponds to one model
+        config variant.
+
+        It is possible for a user to run the profile command with multiple
+        models config variants from different models with potentially different
+        backend parameters. This test includes backend parameters from two
+        separate models, showing that for one particular row (for a 'model A'
+        config variant), it only populates the backend parameter cells for
+        'model A', and the backend parameter cells for 'model B' are empty
+        (None).
+
+        Here is an example table with backend parameters from different models:
+
+        Models (Inference):
+        Model       Model Config Path   backend_parameter/add_sub_key_1   backend_parameter/add_sub_key_2   backend_parameter/add_sub_2_key_1   backend_parameter/add_sub_2_key_2
+        add_sub     add_sub_config_2    add_sub_value_1                   add_sub_value_2                   None                                None
+        add_sub     add_sub_config_0    add_sub_value_1                   add_sub_value_2                   None                                None
+        add_sub     add_sub_config_1    add_sub_value_1                   add_sub_value_2                   None                                None
+        add_sub_2   add_sub_2_config_2  None                              None                              add_sub_2_value_1                   add_sub_2_value_2
+        add_sub_2   add_sub_2_config_1  None                              None                              add_sub_2_value_1                   add_sub_2_value_2
+        add_sub_2   add_sub_2_config_0  None                              None                              add_sub_2_value_1                   add_sub_2_value_2
+        """
+
+        args = [
+            "model-analyzer",
+            "profile",
+            "-f",
+            "config.yml",
+            "--model-repository",
+            ".",
+        ]
+        yaml_str = """
+            profile_models: profile_models
+            inference_output_fields: model_name,batch_size,backend_parameter/model_1_key_1,backend_parameter/model_1_key_2,backend_parameter/model_2_key_1
+        """
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+        state_manager = AnalyzerStateManager(config=config, server=None)
+        result_manager = ResultManager(
+            config=config,
+            state_manager=state_manager,
+            constraint_manager=ConstraintManager(config),
+        )
+        table_manager = ResultTableManager(config=config, result_manager=result_manager)
+        model_config_str = """
+            parameters: {
+            key: "model_1_key_1"
+                value: {
+                string_value:"model_1_value_1"
+                }
+            }
+            parameters: {
+            key:"model_1_key_2"
+                value: {
+                string_value:"model_1_value_2"
+                }
+            }
+            """
+        backend_parameters = text_format.Parse(
+            model_config_str, model_config_pb2.ModelConfig()
+        ).parameters
+        row = table_manager._get_common_row_items(
+            fields=[
+                "model_name",
+                "batch_size",
+                "backend_parameter/model_1_key_1",
+                "backend_parameter/model_1_key_2",
+                "backend_parameter/model_2_key_1",
+            ],
+            batch_sizes="batch_size",
+            concurrencies=None,
+            request_rates=None,
+            satisfies=None,
+            model_name="model_name",
+            model_config_path=None,
+            dynamic_batchings=None,
+            instance_groups=None,
+            max_batch_sizes=None,
+            backend_parameters=backend_parameters,
+        )
+        self.assertEqual(
+            row,
+            ["model_name", "batch_size", "model_1_value_1", "model_1_value_2", None],
+        )
+
+    def _create_single_model_result_table_manager(self):
+        result_manager, config = load_single_model_result_manager()
+
+        return ResultTableManager(config, result_manager)
+
+    def _create_multi_model_result_table_manager(self):
+        result_manager, config = load_multi_model_result_manager()
+
+        return ResultTableManager(config, result_manager)
+
+    def _create_ensemble_result_table_manager(self):
+        result_manager, config = load_ensemble_result_manager()
+
+        return ResultTableManager(config, result_manager)
+
+    def _create_bls_result_table_manager(self):
+        result_manager, config = load_bls_result_manager()
+
+        return ResultTableManager(config, result_manager)
+
+    def _create_request_rate_result_table_manager(self):
+        result_manager, config = load_request_rate_result_manager()
+
+        return ResultTableManager(config, result_manager)
+
+    def tearDown(self):
+        patch.stopall()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_results.py b/src/c++/perf_analyzer/tests/test_results.py
new file mode 100755
index 000000000..2fc45162b
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_results.py
@@ -0,0 +1,248 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.result.results import Results
+
+from .common import test_result_collector as trc
+from .common.test_utils import construct_run_config
+
+
+class TestResults(trc.TestResultCollector):
+    def setUp(self):
+        self._construct_results()
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_contains_model(self):
+        """
+        Test for the existence of expected models
+        """
+        self.assertTrue(self._result.contains_model("modelA"))
+        self.assertTrue(self._result.contains_model("modelB"))
+        self.assertFalse(self._result.contains_model("modelC"))
+
+    def test_contains_model_variant(self):
+        """
+        Test for the existence of expected model variants in modelA
+        """
+        self.assertTrue(self._result.contains_model_variant("modelA", "model_config_0"))
+        self.assertTrue(self._result.contains_model_variant("modelA", "model_config_1"))
+        self.assertTrue(self._result.contains_model_variant("modelA", "model_config_2"))
+        self.assertFalse(
+            self._result.contains_model_variant("modelA", "model_config_3")
+        )
+        self.assertFalse(
+            self._result.contains_model_variant("modelC", "model_config_0")
+        )
+
+    def test_get_list_of_models(self):
+        """
+        Test that the list of models is returned correctly
+        """
+        model_list = self._result.get_list_of_models()
+
+        self.assertEqual(model_list, ["modelA", "modelB"])
+
+    def test_get_list_of_model_config_measurement_tuples(self):
+        """
+        Test that the correct number of measurements is returned per model config
+        """
+        model_config_measurements_list = (
+            self._result.get_list_of_model_config_measurement_tuples()
+        )
+        num_model_config_measurements = [
+            len(mcm) for mcm in model_config_measurements_list
+        ]
+
+        self.assertEqual(num_model_config_measurements[0], 3)
+        self.assertEqual(num_model_config_measurements[1], 2)
+
+    def test_get_list_of_run_config_measurements(self):
+        """
+        Test that the measurements were correctly added to the dictionary
+        """
+        measurements = self._result.get_list_of_run_config_measurements()
+
+        self.assertEqual(measurements, self._measurements_added)
+
+    def test_get_model_measurements_dict(self):
+        """
+        Test that the measurements were correctly added to the model dictionaries
+        """
+        model_measurements = self._result.get_model_measurements_dict("modelA")
+
+        for index, (run_config, measurements) in enumerate(model_measurements.values()):
+            # There were 3 runs per model config. Make sure their values all match
+            base_index = index * 3
+            self.assertEqual(
+                run_config.model_run_configs()[0]
+                .model_config_variant()
+                .model_config.get_config(),
+                self._run_configs[base_index]
+                .model_run_configs()[0]
+                .model_config_variant()
+                .model_config.get_config(),
+            )
+            self.assertEqual(
+                run_config.model_run_configs()[0]
+                .model_config_variant()
+                .model_config.get_config(),
+                self._run_configs[base_index + 1]
+                .model_run_configs()[0]
+                .model_config_variant()
+                .model_config.get_config(),
+            )
+            self.assertEqual(
+                run_config.model_run_configs()[0]
+                .model_config_variant()
+                .model_config.get_config(),
+                self._run_configs[base_index + 2]
+                .model_run_configs()[0]
+                .model_config_variant()
+                .model_config.get_config(),
+            )
+            self.assertEqual(measurements, self._measurements[index])
+
+    def test_bad_get_model_measurements_dict(self):
+        """
+        Test that we return an empty dict and don't assert if the model
+        doesn't exist in the measurements
+        """
+        model_measurements = self._result.get_model_measurements_dict("bad_model_name")
+        self.assertEqual(model_measurements, {})
+
+    def test_get_model_config_measurements_dict(self):
+        """
+        Test that the measurements were correctly added to the model config dictionaries
+        """
+        model_config_measurements = self._result.get_model_variants_measurements_dict(
+            "modelA", "model_config_1"
+        )
+
+        self.assertEqual(model_config_measurements, self._measurements[1])
+
+    def test_get_all_model_variant_measurements(self):
+        """
+        Test that the list of measurements were correctly generated for a
+        given model + model_config combination
+        """
+        run_config, measurements = self._result.get_all_model_variant_measurements(
+            "modelA", "model_config_1"
+        )
+
+        self.assertEqual(
+            run_config.model_run_configs()[0].model_config(),
+            self._run_configs[3]
+            .model_run_configs()[0]
+            .model_config_variant()
+            .model_config,
+        )
+        self.assertEqual(measurements, list(self._measurements[1].values()))
+
+    def _construct_results(self):
+        """
+        Creates an instance of the results class with two models, and
+        various model configs. Treating the measurements as strings,
+        which makes debugging easier
+        """
+        self._result = Results()
+
+        self._run_configs = [
+            construct_run_config("modelA", "model_config_0", "key_A"),
+            construct_run_config("modelA", "model_config_0", "key_B"),
+            construct_run_config("modelA", "model_config_0", "key_C"),
+            construct_run_config("modelA", "model_config_1", "key_D"),
+            construct_run_config("modelA", "model_config_1", "key_E"),
+            construct_run_config("modelA", "model_config_1", "key_F"),
+            construct_run_config("modelA", "model_config_2", "key_G"),
+            construct_run_config("modelA", "model_config_2", "key_H"),
+            construct_run_config("modelA", "model_config_2", "key_I"),
+        ]
+
+        self._measurements = []
+        self._measurements.append(
+            {
+                "model_config_0 -m key_A": "1",
+                "model_config_0 -m key_B": "2",
+                "model_config_0 -m key_C": "3",
+            }
+        )
+        self._measurements.append(
+            {
+                "model_config_1 -m key_D": "4",
+                "model_config_1 -m key_E": "5",
+                "model_config_1 -m key_F": "6",
+            }
+        )
+        self._measurements.append(
+            {
+                "model_config_2 -m key_G": "7",
+                "model_config_2 -m key_H": "8",
+                "model_config_2 -m key_I": "9",
+            }
+        )
+
+        self._result.add_run_config_measurement(self._run_configs[0], "1")
+        self._result.add_run_config_measurement(self._run_configs[1], "2")
+        self._result.add_run_config_measurement(self._run_configs[2], "3")
+
+        self._result.add_run_config_measurement(self._run_configs[3], "4")
+        self._result.add_run_config_measurement(self._run_configs[4], "5")
+        self._result.add_run_config_measurement(self._run_configs[5], "6")
+
+        self._result.add_run_config_measurement(self._run_configs[6], "7")
+        self._result.add_run_config_measurement(self._run_configs[7], "8")
+        self._result.add_run_config_measurement(self._run_configs[8], "9")
+
+        model_run_config_0f = construct_run_config("modelB", "model_config_0", "key_F")
+        model_run_config_0e = construct_run_config("modelB", "model_config_0", "key_E")
+        model_run_config_0d = construct_run_config("modelB", "model_config_0", "key_D")
+        self._result.add_run_config_measurement(model_run_config_0f, "6")
+        self._result.add_run_config_measurement(model_run_config_0e, "5")
+        self._result.add_run_config_measurement(model_run_config_0d, "4")
+
+        model_run_config_1c = construct_run_config("modelB", "model_config_1", "key_C")
+        model_run_config_1b = construct_run_config("modelB", "model_config_1", "key_B")
+        model_run_config_1a = construct_run_config("modelB", "model_config_1", "key_A")
+        self._result.add_run_config_measurement(model_run_config_1c, "3")
+        self._result.add_run_config_measurement(model_run_config_1b, "2")
+        self._result.add_run_config_measurement(model_run_config_1a, "1")
+
+        self._measurements_added = [
+            "1",
+            "2",
+            "3",
+            "4",
+            "5",
+            "6",
+            "7",
+            "8",
+            "9",
+            "6",
+            "5",
+            "4",
+            "3",
+            "2",
+            "1",
+        ]
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_run_config.py b/src/c++/perf_analyzer/tests/test_run_config.py
new file mode 100755
index 000000000..03dc53c9f
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_run_config.py
@@ -0,0 +1,212 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.run.model_run_config import ModelRunConfig
+from model_analyzer.config.run.run_config import RunConfig
+from model_analyzer.perf_analyzer.perf_config import PerfAnalyzerConfig
+from model_analyzer.triton.model.model_config import ModelConfig
+from model_analyzer.triton.model.model_config_variant import ModelConfigVariant
+
+from .common import test_result_collector as trc
+
+
+class TestRunConfig(trc.TestResultCollector):
+    def tearDown(self):
+        patch.stopall()
+
+    def test_triton_env(self):
+        """
+        Test triton env initialized correctly and returns correctly from function
+        """
+        fake_env = {"a": 5, "b": {"c": 7}}
+        rc = RunConfig(fake_env)
+        self.assertEqual(rc.triton_environment(), fake_env)
+
+    def test_model_run_configs(self):
+        """
+        Test adding and getting ModelRunConfigs
+        """
+        mrc1 = ModelRunConfig("model1", MagicMock(), MagicMock())
+        mrc2 = ModelRunConfig("model2", MagicMock(), MagicMock())
+        rc = RunConfig({})
+        rc.add_model_run_config(mrc1)
+        rc.add_model_run_config(mrc2)
+
+        mrc_out = rc.model_run_configs()
+
+        self.assertEqual(mrc_out[0], mrc1)
+        self.assertEqual(mrc_out[1], mrc2)
+
+    def test_representation(self):
+        """
+        Test that representation() is just a string join of member ModelRunConfig's representations
+        """
+        pc1 = PerfAnalyzerConfig()
+        pc1.update_config({"model-name": "TestModel1"})
+        pc2 = PerfAnalyzerConfig()
+        pc2.update_config({"model-name": "TestModel2"})
+        mrc1 = ModelRunConfig(
+            "model1", ModelConfigVariant(MagicMock(), "model1_config_0"), pc1
+        )
+        mrc2 = ModelRunConfig(
+            "model2", ModelConfigVariant(MagicMock(), "model2_config_0"), pc2
+        )
+        rc = RunConfig({})
+        rc.add_model_run_config(mrc1)
+        rc.add_model_run_config(mrc2)
+
+        expected_representation = (
+            "model1_config_0 "
+            + pc1.representation()
+            + "model2_config_0 "
+            + pc2.representation()
+        )
+        self.assertEqual(rc.representation(), expected_representation)
+
+    def test_representation_mrc_removal(self):
+        """
+        Test that representation removes measurement request count
+        """
+        pc = PerfAnalyzerConfig()
+        pc.update_config({"model-name": "TestModel1"})
+        pc.update_config({"measurement-request-count": "500"})
+
+        mrc = ModelRunConfig(
+            "model1", ModelConfigVariant(MagicMock(), "model1_config_0"), pc
+        )
+
+        expected_representation = "model1_config_0 -m TestModel1"
+        self.assertEqual(mrc.representation(), expected_representation)
+
+    def test_cpu_only(self):
+        """
+        Test that cpu_only() is only true if all ModelConfigs are cpu_only()
+        """
+        cpu_only_true_mc = ModelConfig({})
+        cpu_only_true_mcv = ModelConfigVariant(
+            cpu_only_true_mc, "cpu_only_true_config_default", True
+        )
+
+        cpu_only_false_mc = ModelConfig({})
+        cpu_only_false_mcv = ModelConfigVariant(
+            cpu_only_false_mc, "cpu_only_false_config_default", False
+        )
+
+        mrc1 = ModelRunConfig("model1", cpu_only_true_mcv, MagicMock())
+        mrc2 = ModelRunConfig("model2", cpu_only_false_mcv, MagicMock())
+
+        rc = RunConfig({})
+        rc.add_model_run_config(mrc1)
+        self.assertTrue(rc.cpu_only())
+
+        rc.add_model_run_config(mrc2)
+        self.assertFalse(rc.cpu_only())
+
+    def test_mrc_with_illegal_combinations(self):
+        """
+        Test ModelRunConfig with illegal combinations
+        """
+        mc = ModelConfig({})
+        pc = PerfAnalyzerConfig()
+
+        # Valid client batch-size and invalid model preferred_batch_size
+        pc["batch-size"] = 2
+        mc.set_config(
+            {
+                "name": "test_model",
+                "max_batch_size": 4,
+                "dynamic_batching": {"preferred_batch_size": [4, 8]},
+            }
+        )
+        mcv = ModelConfigVariant(mc, "test_model_config_default")
+        mrc = ModelRunConfig("modelA", mcv, pc)
+        self.assertFalse(mrc.is_legal_combination())
+
+        # Invalid client batch-size and invalid model preferred_batch_size
+        pc["batch-size"] = 8
+        mrc = ModelRunConfig("modelB", mcv, pc)
+        self.assertFalse(mrc.is_legal_combination())
+
+        # Invalid client batch-size and valid model preferred_batch_size
+        mc.set_config(
+            {"max_batch_size": 4, "dynamic_batching": {"preferred_batch_size": [2]}}
+        )
+        mrc = ModelRunConfig("modelC", mcv, pc)
+        self.assertFalse(mrc.is_legal_combination())
+
+    def test_mrc_with_legal_combinations(self):
+        """
+        Test ModelRunConfig with legal combinations
+        """
+        mc = ModelConfig({})
+        pc = PerfAnalyzerConfig()
+
+        # Valid client batch-size and no model preferred_batch_size
+        pc["batch-size"] = 2
+        mc.set_config({"max_batch_size": 8})
+        mcv = ModelConfigVariant(mc, "test_model_config_default")
+        mrc = ModelRunConfig("modelA", mcv, pc)
+        self.assertTrue(mrc.is_legal_combination())
+
+        # Valid client batch-size and valid model preferred_batch_size
+        mc.set_config(
+            {"max_batch_size": 8, "dynamic_batching": {"preferred_batch_size": [4]}}
+        )
+        mrc = ModelRunConfig("modelB", mcv, pc)
+        self.assertTrue(mrc.is_legal_combination())
+
+    def test_composing_mrc_with_illegal_combinations(self):
+        """
+        Test ModelRunConfig with illegal combinations in composing config
+        """
+        mc = ModelConfig({})
+        pc = PerfAnalyzerConfig()
+        composing_model_config_variants = [
+            ModelConfigVariant(ModelConfig({}), "composing_model_A_config_default"),
+            ModelConfigVariant(ModelConfig({}), "composing_model_B_config_default"),
+        ]
+
+        # Invalid client batch-size and valid model preferred_batch_size for composing_config[1]
+        pc["batch-size"] = 2
+        mc.set_config({"max_batch_size": 8, "name": "test_model"})
+        mcv = ModelConfigVariant(mc, "test_model_config_default")
+        mrc = ModelRunConfig("modelC", mcv, pc)
+
+        composing_model_config_variants[0].model_config.set_config(
+            {
+                "name": "composing_model_A",
+                "max_batch_size": 4,
+                "dynamic_batching": {"preferred_batch_size": [2]},
+            }
+        )
+        composing_model_config_variants[1].model_config.set_config(
+            {
+                "name": "composing_model_B",
+                "max_batch_size": 4,
+                "dynamic_batching": {"preferred_batch_size": [4, 8]},
+            }
+        )
+
+        mrc.add_composing_model_config_variants(composing_model_config_variants)
+
+        self.assertFalse(mrc.is_legal_combination())
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_run_config_generator.py b/src/c++/perf_analyzer/tests/test_run_config_generator.py
new file mode 100755
index 000000000..b58850b9f
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_run_config_generator.py
@@ -0,0 +1,729 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.generate.brute_run_config_generator import (
+    BruteRunConfigGenerator,
+)
+from model_analyzer.config.generate.generator_utils import GeneratorUtils as utils
+from model_analyzer.config.generate.model_profile_spec import ModelProfileSpec
+from model_analyzer.config.generate.model_run_config_generator import (
+    ModelRunConfigGenerator,
+)
+from model_analyzer.config.generate.model_variant_name_manager import (
+    ModelVariantNameManager,
+)
+from model_analyzer.config.input.config_defaults import (
+    DEFAULT_RUN_CONFIG_MAX_CONCURRENCY,
+    DEFAULT_RUN_CONFIG_MAX_INSTANCE_COUNT,
+    DEFAULT_RUN_CONFIG_MAX_MODEL_BATCH_SIZE,
+)
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+
+from .common import test_result_collector as trc
+from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
+from .mocks.mock_model_config import MockModelConfig
+from .mocks.mock_os import MockOSMethods
+
+
+class TestRunConfigGenerator(trc.TestResultCollector):
+    def __init__(self, methodname):
+        super().__init__(methodname)
+        self._fake_throughput = 1
+
+    def test_default_config_single_model(self):
+        """
+        Test Default Single Model:
+
+        num_PAC = log2(DEFAULT_RUN_CONFIG_MAX_CONCURRENCY) + 1
+        num_MC = (  DEFAULT_RUN_CONFIG_MAX_INSTANCE_COUNT
+                  x log2(DEFAULT_RUN_CONFIG_MAX_MODEL_BATCH_SIZE)
+                 ) + 1
+        total = (num_PAC * num_MC) will be generated by the auto-search
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            profile_models:
+                - my-model
+            """)
+        # yapf: enable
+
+        expected_pa_configs = len(
+            utils.generate_doubled_list(1, DEFAULT_RUN_CONFIG_MAX_CONCURRENCY)
+        )
+
+        expected_model_configs = (
+            DEFAULT_RUN_CONFIG_MAX_INSTANCE_COUNT
+            * len(
+                utils.generate_doubled_list(1, DEFAULT_RUN_CONFIG_MAX_MODEL_BATCH_SIZE)
+            )
+            + 1
+        )
+        expected_num_of_configs = expected_pa_configs * expected_model_configs
+
+        self._run_and_test_run_config_generator(
+            yaml_str, expected_config_count=expected_num_of_configs
+        )
+
+    def test_two_models(self):
+        """
+        Test Two Models that have the same automatic configuration:
+
+        num_PAC = 2
+        num_MC = 2 * 2 = 4
+        model_total = (2 * 4) = 8 will be generated by the auto-search for each model
+
+        default_step = (2 * 2) = 4 will be generated by combining the default configs (2 for each PAC)
+
+        total = default_step + model_total * model_total = 68
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 2
+            run_config_search_max_instance_count: 2
+            run_config_search_max_concurrency: 2
+            profile_models:
+                - my-model
+                - my-modelB
+
+            """)
+        # yapf: enable
+
+        expected_num_of_configs = 68
+
+        # Expect 78 calls to ModelRunConfigGenerator.set_last_results
+        # All 68 experiments will be passed to the leaf generator
+        # All 2 times that the leaf generator is done with default config will also pass results to root generator
+        # All 8 times that the leaf generator is done with non-default config will also pass results to root generator
+
+        expected_num_calls_to_set_last_results = 78
+
+        with patch.object(
+            ModelRunConfigGenerator,
+            "set_last_results",
+            side_effect=ModelRunConfigGenerator.set_last_results,
+            autospec=True,
+        ) as mock_method:
+            self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+            self.assertEqual(
+                mock_method.call_count, expected_num_calls_to_set_last_results
+            )
+
+    def test_two_uneven_models(self):
+        """
+        Test Two Uneven Models:
+
+        Model A is auto search:
+            num_PAC = 3
+            num_MC = 2 * 2 = 4
+            modelA_total = (3 * 4) = 12
+
+        Model B is manual search:
+            num_PAC = 2
+            num_MC = 2 * 3 = 6
+            modelB_total = 2 * 6 = 12
+
+        default_step = num_PAC_A * num_PAC_B = 6
+        total = default_step + modelA_total * modelB_total = 150
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 2
+            run_config_search_max_instance_count: 2
+            run_config_search_max_concurrency: 2
+            profile_models:
+                my-model:
+                    parameters:
+                        concurrency: [1,2,3]
+                my-modelB:
+                    model_config_parameters:
+                        max_batch_size: [1,4,16]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1,2]
+            """)
+        # yapf: enable
+
+        expected_num_of_configs = 150
+
+        self._run_and_test_run_config_generator(
+            yaml_str, expected_config_count=expected_num_of_configs
+        )
+
+    def test_three_uneven_models(self):
+        """
+        Test Three Uneven Models:
+
+        Model A is auto search:
+            num_PAC = 2
+            num_MC = 2 * 2 = 4
+            modelA_total = 2 * 4 = 8
+
+        Model B is auto search with fixed concurrency:
+            num_PAC = 3
+            num_MC = 2 * 2 = 4
+            modelB_total = 3 *4 = 12
+
+        Model C is manual search:
+            num_PAC = 2
+            num_MC = 3 * 2 = 6
+            modelC_total = 2 * 6 = 12
+
+        default_step = num_PAC_A * num_PAC_B * num_PAC_C = 2*3*2 = 12
+
+        total = default_step + modelA_total * modelB_total * modelC_total = 1164
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 2
+            run_config_search_max_instance_count: 2
+            run_config_search_max_concurrency: 2
+            profile_models:
+                - my-model
+                -
+                  my-modelB:
+                    parameters:
+                        concurrency: [1,5,9]
+                -
+                  my-model3:
+                    parameters:
+                        concurrency: [10, 20]
+                    model_config_parameters:
+                        max_batch_size: [1,64]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1,2,3]
+            """)
+        # yapf: enable
+
+        expected_num_of_configs = 1164
+        # All 2 times that the leaf generator is done with default config will also pass results to root generator
+        # All 8 times that the leaf generator is done with non-default config will also pass results to root generator
+
+        # Expect 1268 calls to ModelRunConfigGenerator.set_last_results
+        # All 1164 experiments will be passed to the leaf generator
+        # All 96 times that the leaf generator is done with non-default config will pass results to the middle generator
+        # All 6 times that the leaf generator is done with default config will pass results to the middle generator
+        # All 8 times that the middle generator is done with non-default config will pass results to the root generator
+        # All 2 times that the middle generator is done with default config will pass results to the root generator
+        expected_num_calls_to_set_last_results = 1276
+
+        with patch.object(
+            ModelRunConfigGenerator,
+            "set_last_results",
+            side_effect=ModelRunConfigGenerator.set_last_results,
+            autospec=True,
+        ) as mock_method:
+            self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+            self.assertEqual(
+                mock_method.call_count, expected_num_calls_to_set_last_results
+            )
+
+    def test_early_backoff_leaf_model(self):
+        """
+        Test the case where there are two models, and the 'leaf' model (the last generator called
+        in the recursive generator stack) has early backoff
+
+        Both models are auto search:
+            num_PAC = 2
+            num_MC = 4 * 2 = 8
+            model_total = 2 * 8 = 16
+
+        default_step = num_PAC_A * num_PAC_B = 4
+        total = default_step + modelA_total * modelA_total = 260
+
+        However, the test will set up the throughput values such that ModelB will see a lack
+        of throughput the first time that it is walking max_batch_size.
+        Normally it would walk values 1,2,4,8, but for this ONE case it will only walk 1,2.
+        This will reduce the total count by 4, as there are two concurrencies that would be
+        tried for ModelB for each of the two removed max_batch_sizes
+
+        Thus, actual expected_count = 256
+
+
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 8
+            run_config_search_max_instance_count: 2
+            run_config_search_max_concurrency: 2
+            profile_models:
+                - my-model
+                - my-modelB
+            """)
+        # yapf: enable
+
+        expected_num_of_configs = 256
+
+        perf_throughput_values = [2**i for i in range(expected_num_of_configs)]
+        perf_throughput_values[6] = perf_throughput_values[5]
+        perf_throughput_values[7] = perf_throughput_values[5]
+
+        with patch.object(
+            TestRunConfigGenerator, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = perf_throughput_values
+            self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+    def test_early_backoff_root_model(self):
+        """
+        Test the case where there are two models, and the 'root' model (the first one
+        called in the recursive generator stack) has early backoff
+
+        Both models are auto search:
+            num_PAC = 2
+            num_MC = 4 * 2 = 8
+            model_total = 2 * 8 = 16
+
+        default_step = num_PAC_A * num_PAC_B = 4
+        total = default_step + modelA_total * modelA_total = 260
+
+        However, the test will set up the throughput values such that ModelA will see a lack
+        of throughput the first time that it is walking max_batch_size.
+        Normally it would walk values 1,2,4,8, but for this case it will only walk 1,2.
+        This will reduce the total count by 64: (2 max_batch_sizes * 2 concurrencies * 16 modelB cases)
+
+        Thus, actual expected_count = 196
+
+
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 8
+            run_config_search_max_instance_count: 2
+            run_config_search_max_concurrency: 2
+            profile_models:
+                - my-model
+                - my-modelB
+            """)
+        # yapf: enable
+
+        expected_num_of_configs = 196
+
+        perf_throughput_values = [2**i for i in range(expected_num_of_configs)]
+        # First 4 is modelA=default, modelB=default
+        # Next 32 is modelA max_batch_size=1, modelB=all 16 cases
+        # Next 32 is modelA max_batch_size=2, modelB=all 16 cases. We want to change these to show no increase
+        for i in range(36, 68):
+            perf_throughput_values[i] = perf_throughput_values[i - 36]
+
+        with patch.object(
+            TestRunConfigGenerator, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = perf_throughput_values
+            self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+    def test_measurement_list(self):
+        """
+        Test that the root model (the first one called in the recursive generator stack) gets a list
+        of all measurements since the last time it took a step, and makes a decision based on the
+        maximum throughput observed, not just the last measurement
+
+        Both models are auto search:
+            num_PAC = 2
+            num_MC = 4 * 2 = 8
+            model_total = 2 * 8 = 16
+
+        default_total = num_PAC_A * num_PAC_B = 2*2 = 4
+        total = default_total + modelA_total * modelB_total = 324
+
+        The test will set up the throughput values such that there is an increase to the maximum throughput
+        between each step of the root model, but that those results aren't strictly increasing every time.
+
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 8
+            run_config_search_max_instance_count: 2
+            run_config_search_max_concurrency: 2
+            profile_models:
+                - my-model
+                - my-modelB
+            """)
+        # yapf: enable
+
+        expected_num_of_configs = 260
+
+        # The first 4 results (0-3) are modelA=default, modelB=default
+        # The next 32 results (4-35) are modelA= max_batch_size=1 concurrency 1 and 2, modelB all 16 combinations
+        # The next 32 results (36-67) are modelA= max_batch_size=2 concurrency 1 and 2, modelB all 16 combinations
+        #
+        # Result 67 is the one to change to a smaller result for this test to work. If modelA was incorrectly
+        # only looking at 67 instead of all results 36-67, then it would incorrectly cut off the max_batch_size
+        # search and would not end up returning a full 260 configs
+        perf_throughput_values = list(range(260))
+        perf_throughput_values[67] = 1
+
+        with patch.object(
+            TestRunConfigGenerator, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = perf_throughput_values
+            self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+    def test_matching_triton_server_env(self):
+        """
+        Test that we don't assert if triton server environments match:
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 2
+            run_config_search_max_instance_count: 2
+            run_config_search_max_concurrency: 2
+            profile_models:
+                -
+                  my-model:
+                    triton_server_environment:
+                        'LD_PRELOAD': fake_preload_1,
+                        'LD_LIBRARY_PATH': fake_library_path_1
+                -
+                  my-modelB:
+                    triton_server_environment:
+                        'LD_PRELOAD': fake_preload_1,
+                        'LD_LIBRARY_PATH': fake_library_path_1
+            """)
+        # yapf: enable
+
+        expected_num_of_configs = 68
+        self._run_and_test_run_config_generator(
+            yaml_str, expected_config_count=expected_num_of_configs
+        )
+
+    def test_mismatching_triton_server_env(self):
+        """
+        Test that we assert if triton server environments don't match:
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 2
+            run_config_search_max_instance_count: 2
+            run_config_search_max_concurrency: 2
+            profile_models:
+                -
+                  my-model:
+                    triton_server_environment:
+                        'LD_PRELOAD': fake_preload_1,
+                        'LD_LIBRARY_PATH': fake_library_path_1
+                -
+                  my-modelB:
+                    triton_server_environment:
+                        'LD_PRELOAD': fake_preload_2,
+                        'LD_LIBRARY_PATH': fake_library_path_2
+            """)
+        # yapf: enable
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            expected_num_of_configs = 100
+            self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+    def test_none_result_in_max_batch_size(self):
+        """
+        Test that if we get None results for an entire set of MaxBatch experiments that
+        we properly stop stepping max batch size and do not assert
+
+        When the test steps to 2 instances with max_batch_size of 2, it will return None as the results.
+        This should stop the entire max_batch_size walk for 2 instances, and should step to 3 instances
+        """
+
+        expected_num_of_configs = 17
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 4
+            run_config_search_max_instance_count: 3
+            run_config_search_max_concurrency: 2
+            profile_models:
+                - my-model
+            """)
+
+        perf_throughput_values = [
+            1, 2,  # Default, Concurrency 1 and 2
+            1, 2,  # Instances=1, MaxBatch=1, Concurrency 1 and 2
+            3, 4,  # Instances=1, MaxBatch=2, Concurrency 1 and 2
+            5, 6,  # Instances=1, MaxBatch=4, Concurrency 1 and 2
+            7, 8,  # Instances=2, MaxBatch=1, Concurrency 1 and 2
+            None,  # Instances=2, MaxBatch=2, Concurrency 1.
+            9,10,  # Instances=3, MaxBatch=1, Concurrency 1 and 2
+            11,12, # Instances=3, MaxBatch=2, Concurrency 1 and 2
+            13,14  # Instances=3, MaxBatch=4, Concurrency 1 and 2
+        ]
+        # yapf: enable
+
+        with patch.object(
+            TestRunConfigGenerator, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = perf_throughput_values
+            self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+    def test_none_result_before_threshold(self):
+        """
+        Test that a 'none' result (from PA erroring) before the minimum number
+        of throughput runs (4) will properly early exit and not assert
+        """
+
+        # Full sweep is 3 model configs * 12 concurrencies = 36.
+        # However, we will early exit and skip 9 of the concurrencies
+        # Thus, expect 27
+        #
+        expected_num_of_configs = 27
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 2
+            run_config_search_max_instance_count: 1
+            run_config_search_max_concurrency: 2048
+            profile_models:
+                - my-model
+            """)
+
+        perf_throughput_values = [
+            1,2,4,8,16,32,64,128,256,512,1024,2048, # Default config
+            1,2,None,                               # Batch size 1
+            1,2,4,8,16,32,64,128,256,512,1024,2048  # Batch size 2
+        ]
+        # yapf: enable
+
+        with patch.object(
+            TestRunConfigGenerator, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = perf_throughput_values
+            self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+    def test_none_result_after_threshold(self):
+        """
+        Test that a 'none' result (from PA erroring) after the minimum number
+        of throughput runs (4) will properly early exit and not assert
+        """
+
+        # Full sweep is 3 model configs * 12 concurrencies = 36.
+        # However, we will early exit and skip 1 of the concurrencies
+        # Thus, expect 35
+        #
+        expected_num_of_configs = 35
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 2
+            run_config_search_max_instance_count: 1
+            run_config_search_max_concurrency: 2048
+            profile_models:
+                - my-model
+            """)
+
+        perf_throughput_values = [
+            1,2,4,8,16,32,64,128,256,512,1024,2048, # Default config
+            1,2,4,8,16,32,64,128,256,512,None,      # Batch size 1
+            1,2,4,8,16,32,64,128,256,512,1024,2048  # Batch size 2
+        ]
+        # yapf: enable
+
+        with patch.object(
+            TestRunConfigGenerator, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = perf_throughput_values
+            self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+    def test_variant_naming(self):
+        """
+        Confirm that model variant names are consistent
+
+        The test is set up such that on the first pass, we will skip
+        model B's case of 1 instance, batch size=4. On the second pass
+        (when the root model max_batch_size of 2), no cases will be skipped.
+        As a result, the naming will be this for model B:
+            modelB_config_0: 1 instance, max_batch_size=1
+            modelB_config_1: 1 instance, max_batch_size=2
+            modelB_config_5: 1 instance, max_batch_size=4
+            modelB_config_2: 2 instance, max_batch_size=1
+            modelB_config_3: 2 instance, max_batch_size=2
+            modelB_config_4: 2 instance, max_batch_size=4
+        """
+
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_model_batch_size: 4
+            run_config_search_max_instance_count: 2
+            run_config_search_max_concurrency: 1
+
+            profile_models:
+                -
+                  modelA:
+                    model_config_parameters:
+                        max_batch_size: [1,2]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1]
+                - modelB
+            """)
+
+        expected_modelB_name_order = [
+            "modelB_config_default",
+            "modelB_config_0",
+            "modelB_config_1",
+            "modelB_config_2",
+            "modelB_config_3",
+            "modelB_config_4",
+            "modelB_config_0",
+            "modelB_config_1",
+            "modelB_config_5", # This was skipped, and needs to be created on second pass
+            "modelB_config_2",
+            "modelB_config_3",
+            "modelB_config_4",
+        ]
+
+        perf_throughput_values = [
+            1,      # Default config
+            1,None, # A: 1 Instance, BS=1  B: 1 Instance, BS=1,2
+            2,3,4,  # A: 1 Instance, BS=1  B: 2 Instance, BS=1,2,4
+            5,6,7,  # A: 1 Instance, BS=2  B: 1 Instance, BS=1,2,4
+            8,9,10  # A: 1 Instance, BS=2  B: 2 Instance, BS=1,2,4
+        ]
+        # yapf: enable
+
+        expected_num_of_configs = 12
+
+        with patch.object(
+            TestRunConfigGenerator, "_get_next_perf_throughput_value"
+        ) as mock_method:
+            mock_method.side_effect = perf_throughput_values
+            run_configs = self._run_and_test_run_config_generator(
+                yaml_str, expected_config_count=expected_num_of_configs
+            )
+
+        for i, rc in enumerate(run_configs):
+            self.assertEqual(
+                expected_modelB_name_order[i],
+                rc.model_run_configs()[1].model_variant_name(),
+            )
+
+    def _run_and_test_run_config_generator(self, yaml_str, expected_config_count):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+        ]
+
+        protobuf = """
+            max_batch_size: 8
+            instance_group [
+            {
+                kind: KIND_CPU
+                count: 1
+            }
+            ]
+            """
+
+        self.mock_model_config = MockModelConfig(protobuf)
+        self.mock_model_config.start()
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        profile_models = []
+        for model in config.profile_models:
+            profile_models.append(
+                ModelProfileSpec(model, config, MagicMock(), MagicMock())
+            )
+
+        rcg = BruteRunConfigGenerator(
+            config, MagicMock(), profile_models, MagicMock(), ModelVariantNameManager()
+        )
+
+        run_configs = []
+        for run_config in rcg.get_configs():
+            run_configs.append(run_config)
+            rcg.set_last_results(self._get_next_fake_results())
+
+        self.assertEqual(expected_config_count, len(set(run_configs)))
+
+        # Checks that each ModelRunConfig contains the expected number of model_configs
+        for run_config in run_configs:
+            self.assertEqual(
+                len(run_config.model_run_configs()), len(config.profile_models)
+            )
+
+        self.mock_model_config.stop()
+
+        return run_configs
+
+    def setUp(self):
+        # Mock path validation
+        self.mock_os = MockOSMethods(
+            mock_paths=["model_analyzer.config.input.config_utils"]
+        )
+        self.mock_os.start()
+
+    def tearDown(self):
+        self.mock_os.stop()
+        patch.stopall()
+
+    def _get_next_fake_results(self):
+        throughput_value = self._get_next_perf_throughput_value()
+
+        measurement = None
+
+        if throughput_value is not None:
+            measurement = construct_run_config_measurement(
+                model_name=MagicMock(),
+                model_config_names=["test_config_name"],
+                model_specific_pa_params=MagicMock(),
+                gpu_metric_values=MagicMock(),
+                non_gpu_metric_values=[{"perf_throughput": throughput_value}],
+            )
+
+        return [measurement]
+
+    def _get_next_perf_throughput_value(self):
+        self._fake_throughput *= 2
+        return self._fake_throughput
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_run_config_measurement.py b/src/c++/perf_analyzer/tests/test_run_config_measurement.py
new file mode 100755
index 000000000..579af9e2a
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_run_config_measurement.py
@@ -0,0 +1,739 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import json
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.result.run_config_measurement import RunConfigMeasurement
+from tests.common.test_utils import (
+    construct_constraint_manager,
+    construct_run_config_measurement,
+    convert_avg_gpu_metrics_to_data,
+    convert_gpu_metrics_to_data,
+    convert_non_gpu_metrics_to_data,
+    default_encode,
+)
+
+from .common import test_result_collector as trc
+
+
+class TestRunConfigMeasurement(trc.TestResultCollector):
+    def setUp(self):
+        self._construct_rcm0()
+        self._construct_rcm1()
+        self._construct_rcm2()
+        self._construct_rcm3()
+        self._construct_rcm4()
+        self._construct_rcm5()
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_model_variants_name(self):
+        """
+        Test that the model_variants_name was initialized correctly
+        """
+        self.assertEqual(self.rcm0.model_variants_name(), self.model_variants_name)
+
+    def test_gpu_data(self):
+        """
+        Test that the gpu data is correct
+        """
+        self.assertEqual(
+            self.rcm0.gpu_data(), convert_gpu_metrics_to_data(self.gpu_metric_values)
+        )
+
+    def test_non_gpu_data(self):
+        """
+        Test that the non-gpu data is correct
+        """
+        self.assertEqual(
+            self.rcm0.non_gpu_data(),
+            [
+                convert_non_gpu_metrics_to_data(ngvm)
+                for ngvm in self.rcm0_non_gpu_metric_values
+            ],
+        )
+
+    def test_data(self):
+        """
+        Test that the gpu + non-gpu data is correct
+        """
+        avg_gpu_data = convert_avg_gpu_metrics_to_data(self.avg_gpu_metric_values)
+
+        data = [
+            list(avg_gpu_data.values()) + convert_non_gpu_metrics_to_data(ngvm)
+            for ngvm in self.rcm0_non_gpu_metric_values
+        ]
+
+        self.assertEqual(list(self.rcm0.data().values()), data)
+
+    def test_gpus_used(self):
+        """
+        Test that the list of gpus used is correct
+        """
+        self.assertEqual(self.rcm0.gpus_used(), ["0", "1"])
+
+    def test_get_gpu_metric(self):
+        """
+        Test that the gpu metric data is correct
+        """
+        avg_gpu_data = convert_avg_gpu_metrics_to_data(self.avg_gpu_metric_values)
+
+        self.assertEqual(
+            self.rcm0.get_gpu_metric("gpu_used_memory"), avg_gpu_data["gpu_used_memory"]
+        )
+
+        self.assertEqual(
+            self.rcm0.get_gpu_metric("gpu_utilization"), avg_gpu_data["gpu_utilization"]
+        )
+
+    def test_gpu_gpu_metric_not_found(self):
+        """
+        Test that an incorrect metric search returns None
+        """
+        self.assertEqual(self.rcm0.get_gpu_metric("XXXXX"), None)
+
+    def test_get_non_gpu_metric(self):
+        """
+        Test that the non-gpu metric data is correct
+        """
+        non_gpu_data = [
+            convert_non_gpu_metrics_to_data(non_gpu_metric_value)
+            for non_gpu_metric_value in self.rcm0_non_gpu_metric_values
+        ]
+
+        self.assertEqual(
+            self.rcm0.get_non_gpu_metric("perf_throughput"),
+            [non_gpu_data[0][0], non_gpu_data[1][0]],
+        )
+        self.assertEqual(
+            self.rcm0.get_non_gpu_metric("perf_latency_p99"),
+            [non_gpu_data[0][1], non_gpu_data[1][1]],
+        )
+        self.assertEqual(
+            self.rcm0.get_non_gpu_metric("cpu_used_ram"),
+            [non_gpu_data[0][2], non_gpu_data[1][2]],
+        )
+
+    def test_get_weighted_non_gpu_metric(self):
+        """
+        Test that the weighted non-gpu metric data is correct
+        """
+        non_gpu_data = [
+            convert_non_gpu_metrics_to_data(weighted_non_gpu_metric_value)
+            for weighted_non_gpu_metric_value in self.rcm0_weighted_non_gpu_metric_values
+        ]
+
+        self.assertEqual(
+            self.rcm0.get_weighted_non_gpu_metric("perf_throughput"),
+            [non_gpu_data[0][0], non_gpu_data[1][0]],
+        )
+        self.assertEqual(
+            self.rcm0.get_weighted_non_gpu_metric("perf_latency_p99"),
+            [non_gpu_data[0][1], non_gpu_data[1][1]],
+        )
+        self.assertEqual(
+            self.rcm0.get_weighted_non_gpu_metric("cpu_used_ram"),
+            [non_gpu_data[0][2], non_gpu_data[1][2]],
+        )
+
+    def test_non_gpu_get_metric_value(self):
+        """
+        Test that the non-gpu metric value is correct
+        """
+        self.assertEqual(
+            self.rcm0.get_non_gpu_metric_value("perf_throughput"),
+            sum(
+                [
+                    self.rcm0_non_gpu_metric_values[0]["perf_throughput"],
+                    self.rcm0_non_gpu_metric_values[1]["perf_throughput"],
+                ]
+            ),
+        )
+
+    def test_gpu_get_metric_value(self):
+        """
+        Test that the gpu metric value is correct
+        """
+        self.assertEqual(
+            self.rcm0.get_gpu_metric_value("gpu_used_memory"),
+            self.avg_gpu_metric_values["gpu_used_memory"],
+        )
+
+    def test_get_weighted_non_gpu_metric_value(self):
+        """
+        Test that the non-gpu weighted metric value is correct
+        """
+        sum_weighted_metric_value = (
+            (self.rcm0_non_gpu_metric_values[0]["perf_latency_p99"] * self.weights[0])
+            + (self.rcm0_non_gpu_metric_values[1]["perf_latency_p99"] * self.weights[1])
+        ) / sum(self.weights)
+
+        mean_weighted_metric_value = sum_weighted_metric_value / len(
+            self.rcm0_weighted_non_gpu_metric_values
+        )
+
+        self.assertEqual(
+            self.rcm0.get_weighted_non_gpu_metric_value("perf_latency_p99"),
+            mean_weighted_metric_value,
+        )
+
+    def test_model_specific_pa_params(self):
+        """
+        Test that the model specific PA params are correct
+        """
+        self.assertEqual(
+            self.rcm0.model_specific_pa_params(), self.model_specific_pa_params
+        )
+
+    def test_is_better_than(self):
+        """
+        Test to ensure measurement comparison is working as intended
+        """
+        # RCM0: 1000, 40    RCM1: 500, 30  weights:[1,3]
+        # RCM0-A's throughput is better than RCM1-A (0.5)
+        # RCM0-B's latency is worse than RCM1-B (-0.25)
+        # Factoring in model config weighting
+        # tips this is favor of RCM1 (0.125, -0.1875)
+        self.assertFalse(self.rcm0.is_better_than(self.rcm1))
+
+        # This tips the scale in the favor of RCM0 (0.2, -0.15)
+        self.rcm0.set_model_config_weighting([2, 3])
+        self.assertTrue(self.rcm0.is_better_than(self.rcm1))
+        self.assertGreater(self.rcm0, self.rcm1)
+
+    def test_is_better_than_consistency(self):
+        """
+        Test to ensure measurement comparison is working correctly
+        when the percentage gain is very close
+        """
+        # Throughputs
+        #   RCM2{A/B}: {336,223}   RCM3{A/B}: {270,272}
+        #   RCM2-A is 21.8% better than RCM3-A
+        #   RMC2-B is 19.8% worse than RCM3-B
+        # Therefore, RCM2 is (very slightly) better than RCM3
+        self.assertTrue(self.rcm2.is_better_than(self.rcm3))
+        self.assertFalse(self.rcm3.is_better_than(self.rcm2))
+
+    def test_compare_measurements(self):
+        """
+        Test to ensure compare measurement function returns
+        the correct magnitude
+        # RCM4's throughput is 1000
+        # RCM5's throughput is 2000
+        # Therefore, the magnitude is (RCM5 - RCM4) / avg throughput
+        #                             (2000 - 1000) / 1500
+        """
+        rcm4_vs_rcm5 = self.rcm4.compare_measurements(self.rcm5)
+        self.assertEqual(rcm4_vs_rcm5, 1000 / 1500)
+
+        rcm5_vs_rcm4 = self.rcm5.compare_measurements(self.rcm4)
+        self.assertEqual(rcm5_vs_rcm4, -1000 / 1500)
+
+    def test_is_passing_constraints_none(self):
+        """
+        Test to ensure constraints are reported as passing
+        if none were specified
+        """
+        self.rcm5.set_constraint_manager(
+            construct_constraint_manager(
+                """
+            profile_models:
+              modelA
+            """
+            )
+        )
+        self.assertTrue(self.rcm5.is_passing_constraints())
+
+    def test_is_passing_constraints(self):
+        """
+        Test to ensure constraints are reported as
+        passing/failing if model is above/below
+        throughput threshold
+        """
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 500
+            """
+        )
+        self.rcm5.set_constraint_manager(constraint_manager)
+
+        self.assertTrue(self.rcm5.is_passing_constraints())
+
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 3000
+            """
+        )
+        self.rcm5.set_constraint_manager(constraint_manager)
+
+        self.assertFalse(self.rcm5.is_passing_constraints())
+
+    def test_compare_constraints_none(self):
+        """
+        Checks case where either self or other is passing constraints
+        """
+        # RCM4's throughput is 1000
+        # RCM5's throughput is 2000
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 500
+            """
+        )
+        self.rcm4.set_constraint_manager(constraint_manager)
+
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 2500
+            """
+        )
+        self.rcm5.set_constraint_manager(constraint_manager)
+
+        self.assertEqual(self.rcm4.compare_constraints(self.rcm5), None)
+        self.assertEqual(self.rcm5.compare_constraints(self.rcm4), None)
+
+    def test_compare_constraints_equal(self):
+        """
+        Test to ensure compare constraints reports zero when both
+        RCMs are missing constraints by the same amount
+        """
+        # RCM4's throughput is 1000
+        # RCM5's throughput is 2000
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 1250
+            """
+        )
+        self.rcm4.set_constraint_manager(constraint_manager)
+
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 2500
+            """
+        )
+        self.rcm5.set_constraint_manager(constraint_manager)
+
+        # RCM4 is failing by 20%, RCM5 is failing by 20%
+        self.assertEqual(self.rcm4.compare_constraints(self.rcm5), 0)
+        self.assertEqual(self.rcm5.compare_constraints(self.rcm4), 0)
+
+    def test_compare_constraints_unequal(self):
+        """
+        Test to ensure compare constraints reports the correct
+        value when the RCMs are both failing constraints by different
+        amounts
+        """
+        # RCM4's throughput is 1000
+        # RCM5's throughput is 2000
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 2000
+            """
+        )
+        self.rcm4.set_constraint_manager(constraint_manager)
+
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_throughput:
+                    min: 2500
+            """
+        )
+        self.rcm5.set_constraint_manager(constraint_manager)
+
+        # RCM4 is failing by 50%, RCM5 is failing by 20%
+        self.assertEqual(self.rcm4.compare_constraints(self.rcm5), 0.30)
+        self.assertEqual(self.rcm5.compare_constraints(self.rcm4), -0.30)
+
+    def test_calculate_weighted_percentage_gain(self):
+        """
+        Test to ensure weighted percentage gain is being calculated correctly
+        """
+
+        # RCM0: 1000, 40    RCM1: 500, 30  weights:[1,3]
+        # RCM0-A's throughput is better than RCM1-A (0.5)
+        # RCM0-B's latency is worse than RCM1-B (-0.25)
+        # Factoring in model config weighting
+        # tips this is favor of RCM1 (0.125, -0.1875)
+        # However, by percentage RCM0 will be evaluated as slightly better
+        # 100% on throughput, -25% on latency
+        # Factoring in weighting, RCM0 is slightly better (100 - 75) / 4 = 6.25%
+        self.assertEqual(self.rcm0.calculate_weighted_percentage_gain(self.rcm1), 6.25)
+
+        # Changing the weighting tips the scale in the favor of RCM0 (0.2, -0.15)
+        # And, from a percentage standpoint we get: (200 - 75) / 5 = 25%
+        self.rcm0.set_model_config_weighting([2, 3])
+        self.assertEqual(self.rcm0.calculate_weighted_percentage_gain(self.rcm1), 25.0)
+
+    def test_from_dict(self):
+        """
+        Test to ensure class can be correctly restored from a dictionary
+        """
+        rcm0_json = json.dumps(self.rcm0, default=default_encode)
+
+        rcm0_from_dict = RunConfigMeasurement.from_dict(json.loads(rcm0_json))
+
+        self.assertEqual(
+            rcm0_from_dict.model_variants_name(), self.rcm0.model_variants_name()
+        )
+        self.assertEqual(rcm0_from_dict.gpu_data(), self.rcm0.gpu_data())
+        self.assertEqual(rcm0_from_dict.non_gpu_data(), self.rcm0.non_gpu_data())
+        self.assertEqual(
+            list(rcm0_from_dict.data().values()), list(self.rcm0.data().values())
+        )
+        self.assertEqual(
+            rcm0_from_dict._model_config_measurements,
+            self.rcm0._model_config_measurements,
+        )
+        self.assertEqual(rcm0_from_dict._model_config_weights, [])
+
+    def _construct_rcm0(self):
+        self.model_name = "modelA,modelB"
+        self.model_config_name = ["modelA_config_0", "modelB_config_1"]
+        self.model_variants_name = "".join(self.model_config_name)
+        self.model_specific_pa_params = [
+            {"batch_size": 1, "concurrency": 1},
+            {"batch_size": 2, "concurrency": 2},
+        ]
+
+        self.gpu_metric_values = {
+            "0": {"gpu_used_memory": 6000, "gpu_utilization": 60},
+            "1": {"gpu_used_memory": 10000, "gpu_utilization": 20},
+        }
+        self.avg_gpu_metric_values = {"gpu_used_memory": 8000, "gpu_utilization": 40}
+
+        self.rcm0_non_gpu_metric_values = [
+            {
+                # modelA_config_0
+                "perf_throughput": 1000,
+                "perf_latency_p99": 20,
+                "cpu_used_ram": 1000,
+            },
+            {
+                # modelB_config_1
+                "perf_throughput": 2000,
+                "perf_latency_p99": 40,
+                "cpu_used_ram": 1500,
+            },
+        ]
+
+        self.metric_objectives = [{"perf_throughput": 1}, {"perf_latency_p99": 1}]
+
+        self.weights = [1, 3]
+
+        self.rcm0_weighted_non_gpu_metric_values = []
+        for index, non_gpu_metric_values in enumerate(self.rcm0_non_gpu_metric_values):
+            self.rcm0_weighted_non_gpu_metric_values.append(
+                {
+                    objective: value * self.weights[index] / sum(self.weights)
+                    for (objective, value) in non_gpu_metric_values.items()
+                }
+            )
+
+        self.rcm0 = construct_run_config_measurement(
+            self.model_name,
+            self.model_config_name,
+            self.model_specific_pa_params,
+            self.gpu_metric_values,
+            self.rcm0_non_gpu_metric_values,
+            MagicMock(),
+            self.metric_objectives,
+            self.weights,
+        )
+
+    def _construct_rcm1(self):
+        model_name = "modelA,modelB"
+        model_config_name = ["modelA_config_2", "modelB_config_3"]
+        model_specific_pa_params = [
+            {"batch_size": 3, "concurrency": 3},
+            {"batch_size": 4, "concurrency": 4},
+        ]
+
+        gpu_metric_values = {
+            "0": {"gpu_used_memory": 7000, "gpu_utilization": 40},
+            "1": {"gpu_used_memory": 12000, "gpu_utilization": 30},
+        }
+
+        self.rcm1_non_gpu_metric_values = [
+            {
+                # modelA_config_2
+                "perf_throughput": 500,
+                "perf_latency_p99": 20,
+                "cpu_used_ram": 1000,
+            },
+            {
+                # modelB_config_3
+                "perf_throughput": 1200,
+                "perf_latency_p99": 30,
+                "cpu_used_ram": 1500,
+            },
+        ]
+
+        metric_objectives = [{"perf_throughput": 1}, {"perf_throughput": 1}]
+
+        weights = [1, 3]
+
+        self.rcm1_weighted_non_gpu_metric_values = []
+        for index, non_gpu_metric_values in enumerate(self.rcm1_non_gpu_metric_values):
+            self.rcm1_weighted_non_gpu_metric_values.append(
+                {
+                    objective: value * self.weights[index] / sum(weights)
+                    for (objective, value) in non_gpu_metric_values.items()
+                }
+            )
+
+        self.rcm1 = construct_run_config_measurement(
+            model_name,
+            model_config_name,
+            model_specific_pa_params,
+            gpu_metric_values,
+            self.rcm1_non_gpu_metric_values,
+            MagicMock(),
+            metric_objectives,
+            weights,
+        )
+
+    def _construct_rcm2(self):
+        model_name = "modelA,modelB"
+        model_config_name = ["modelA_config_1", "modelB_config_2"]
+        model_specific_pa_params = [
+            {"batch_size": 3, "concurrency": 3},
+            {"batch_size": 4, "concurrency": 4},
+        ]
+
+        gpu_metric_values = {
+            "0": {"gpu_used_memory": 7000, "gpu_utilization": 40},
+            "1": {"gpu_used_memory": 12000, "gpu_utilization": 30},
+        }
+
+        self.rcm2_non_gpu_metric_values = [
+            {
+                # modelA_config_1
+                "perf_throughput": 336,
+                "perf_latency_p99": 20,
+                "cpu_used_ram": 1000,
+            },
+            {
+                # modelB_config_2
+                "perf_throughput": 223,
+                "perf_latency_p99": 30,
+                "cpu_used_ram": 1500,
+            },
+        ]
+
+        metric_objectives = [{"perf_throughput": 1}, {"perf_throughput": 1}]
+
+        weights = [1, 1]
+
+        self.rcm2_weighted_non_gpu_metric_values = []
+        for index, non_gpu_metric_values in enumerate(self.rcm2_non_gpu_metric_values):
+            self.rcm2_weighted_non_gpu_metric_values.append(
+                {
+                    objective: value * self.weights[index] / sum(weights)
+                    for (objective, value) in non_gpu_metric_values.items()
+                }
+            )
+
+        self.rcm2 = construct_run_config_measurement(
+            model_name,
+            model_config_name,
+            model_specific_pa_params,
+            gpu_metric_values,
+            self.rcm2_non_gpu_metric_values,
+            MagicMock(),
+            metric_objectives,
+            weights,
+        )
+
+    def _construct_rcm3(self):
+        model_name = "modelA,modelB"
+        model_config_name = ["modelA_config_1", "modelB_config_2"]
+        model_specific_pa_params = [
+            {"batch_size": 3, "concurrency": 3},
+            {"batch_size": 4, "concurrency": 4},
+        ]
+
+        gpu_metric_values = {
+            "0": {"gpu_used_memory": 7000, "gpu_utilization": 40},
+            "1": {"gpu_used_memory": 12000, "gpu_utilization": 30},
+        }
+
+        self.rcm3_non_gpu_metric_values = [
+            {
+                # modelA_config_1
+                "perf_throughput": 270,
+                "perf_latency_p99": 20,
+                "cpu_used_ram": 1000,
+            },
+            {
+                # modelB_config_2
+                "perf_throughput": 272,
+                "perf_latency_p99": 30,
+                "cpu_used_ram": 1500,
+            },
+        ]
+
+        metric_objectives = [{"perf_throughput": 1}, {"perf_throughput": 1}]
+
+        weights = [1, 1]
+
+        self.rcm3_weighted_non_gpu_metric_values = []
+        for index, non_gpu_metric_values in enumerate(self.rcm3_non_gpu_metric_values):
+            self.rcm3_weighted_non_gpu_metric_values.append(
+                {
+                    objective: value * self.weights[index] / sum(weights)
+                    for (objective, value) in non_gpu_metric_values.items()
+                }
+            )
+
+        self.rcm3 = construct_run_config_measurement(
+            model_name,
+            model_config_name,
+            model_specific_pa_params,
+            gpu_metric_values,
+            self.rcm3_non_gpu_metric_values,
+            MagicMock(),
+            metric_objectives,
+            weights,
+        )
+
+    def _construct_rcm4(self):
+        model_name = "modelA"
+        model_config_name = ["modelA_config_0"]
+        model_specific_pa_params = [
+            {"batch_size": 1, "concurrency": 1},
+            {"batch_size": 2, "concurrency": 2},
+        ]
+
+        gpu_metric_values = {
+            "0": {"gpu_used_memory": 6000, "gpu_utilization": 60},
+            "1": {"gpu_used_memory": 10000, "gpu_utilization": 20},
+        }
+
+        self.rcm4_non_gpu_metric_values = [
+            {
+                # modelA_config_0
+                "perf_throughput": 1000,
+                "perf_latency_p99": 20,
+                "cpu_used_ram": 1000,
+            },
+        ]
+
+        metric_objectives = [{"perf_throughput": 1}]
+
+        weights = [1]
+
+        self.rcm4_weighted_non_gpu_metric_values = []
+        for index, non_gpu_metric_values in enumerate(self.rcm4_non_gpu_metric_values):
+            self.rcm4_weighted_non_gpu_metric_values.append(
+                {
+                    objective: value * self.weights[index] / sum(self.weights)
+                    for (objective, value) in non_gpu_metric_values.items()
+                }
+            )
+
+        self.rcm4 = construct_run_config_measurement(
+            model_name,
+            model_config_name,
+            model_specific_pa_params,
+            gpu_metric_values,
+            self.rcm4_non_gpu_metric_values,
+            MagicMock(),
+            metric_objectives,
+            weights,
+        )
+
+    def _construct_rcm5(self):
+        model_name = "modelA"
+        model_config_name = ["modelA_config_0"]
+        model_specific_pa_params = [
+            {"batch_size": 1, "concurrency": 1},
+            {"batch_size": 2, "concurrency": 2},
+        ]
+
+        gpu_metric_values = {
+            "0": {"gpu_used_memory": 6000, "gpu_utilization": 60},
+            "1": {"gpu_used_memory": 10000, "gpu_utilization": 20},
+        }
+
+        self.rcm5_non_gpu_metric_values = [
+            {
+                # modelA_config_0
+                "perf_throughput": 2000,
+                "perf_latency_p99": 20,
+                "cpu_used_ram": 1000,
+            },
+        ]
+
+        metric_objectives = [{"perf_throughput": 1}]
+
+        weights = [1]
+
+        self.rcm5_weighted_non_gpu_metric_values = []
+        for index, non_gpu_metric_values in enumerate(self.rcm5_non_gpu_metric_values):
+            self.rcm5_weighted_non_gpu_metric_values.append(
+                {
+                    objective: value * self.weights[index] / sum(self.weights)
+                    for (objective, value) in non_gpu_metric_values.items()
+                }
+            )
+
+        self.rcm5 = construct_run_config_measurement(
+            model_name,
+            model_config_name,
+            model_specific_pa_params,
+            gpu_metric_values,
+            self.rcm5_non_gpu_metric_values,
+            MagicMock(),
+            metric_objectives,
+            weights,
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_run_config_result.py b/src/c++/perf_analyzer/tests/test_run_config_result.py
new file mode 100755
index 000000000..88b723721
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_run_config_result.py
@@ -0,0 +1,341 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.result.run_config_result import RunConfigResult
+from tests.common.test_utils import (
+    construct_constraint_manager,
+    construct_run_config_measurement,
+    convert_non_gpu_metrics_to_data,
+)
+
+from .common import test_result_collector as trc
+
+
+class TestRunConfigResult(trc.TestResultCollector):
+    def setUp(self):
+        self.default_constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              modelA:
+                constraints:
+                  perf_latency_p99:
+                    max: 100
+              modelB:
+                constraints:
+                  perf_latency_p99:
+                    max: 50
+            """
+        )
+        self._construct_empty_rcr()
+        self._construct_throughput_with_latency_constraint_rcr()
+        self._construct_throughput_with_latency_constraint_multi_model_rcr()
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_model_name(self):
+        """
+        Test that model_name is correctly returned
+        """
+        self.assertEqual(self.rcr_empty.model_name(), self.model_name)
+
+    def test_run_config(self):
+        """
+        Test that run_config is correctly returned
+        """
+        self.assertEqual(self.rcr_empty.run_config(), self.run_config)
+
+    def test_failing_empty(self):
+        """
+        Test that failing returns true if no measurements are added
+        """
+        self.assertTrue(self.rcr_empty.failing())
+
+    def test_failing_true(self):
+        """
+        Test that failing returns true if only failing measurements are present
+        """
+        rcr = self._rcr_throughput_with_latency_constraint
+
+        for i in range(2, 6):
+            self._add_rcm_to_rcr(rcr, throughput_value=10 * i, latency_value=100 * i)
+
+        self.assertTrue(rcr.failing())
+
+    def test_failing_true_multi_model(self):
+        """
+        Test that failing returns true if only failing measurements are present
+        in a multi-model configuration
+        """
+        rcr = self._rcr_throughput_with_latency_constraint_multi_model
+
+        # Model A will have some passing measurements,
+        # Model B will have all failing measurements
+        for i in range(1, 6):
+            self._add_multi_model_rcm_to_rcr(
+                rcr, throughput_values=[5 * i, 20 * i], latency_values=[50 * i, 200 * i]
+            )
+
+        self.assertTrue(rcr.failing())
+
+    def test_failing_false(self):
+        """
+        Test that failing returns false if any passing measurements are present
+        """
+        rcr = self._rcr_throughput_with_latency_constraint
+
+        # This also tests the boundary condition: exact match (100) is passing
+        for i in range(1, 6):
+            self._add_rcm_to_rcr(rcr, throughput_value=10 * i, latency_value=100 * i)
+
+        self.assertFalse(rcr.failing())
+
+    def test_failing_false_multi_model(self):
+        """
+        Test that failing returns false if any passing measurements are present
+        in a multi-model configuration
+        """
+        rcr = self._rcr_throughput_with_latency_constraint_multi_model
+
+        for i in range(1, 6):
+            self._add_multi_model_rcm_to_rcr(
+                rcr, throughput_values=[3 * i, 4 * i], latency_values=[30 * i, 40 * i]
+            )
+
+        self.assertFalse(rcr.failing())
+
+    def test_passing_failing_measurements(self):
+        """
+        Test that passing/failing measurements have the correct number
+        of entries
+        """
+        rcr = self._rcr_throughput_with_latency_constraint
+
+        # 4 passing, 6 failing
+        for i in range(1, 11):
+            self._add_rcm_to_rcr(rcr, throughput_value=10 * i, latency_value=25 * i)
+
+        # The heap is unordered so just checking for correct count
+        self.assertEqual(len(rcr.passing_measurements()), 4)
+        self.assertEqual(len(rcr.failing_measurements()), 6)
+
+    def test_passing_failing_measurements_multi_model(self):
+        """
+        Test that passing/failing measurements have the correct number
+        of entries
+        """
+        rcr = self._rcr_throughput_with_latency_constraint_multi_model
+
+        # 2 passing, 10 failing
+        for i in range(1, 13):
+            self._add_multi_model_rcm_to_rcr(
+                rcr, throughput_values=[10 * i, 15 * i], latency_values=[15 * i, 20 * i]
+            )
+
+        # The heap is unordered so just checking for correct count
+        self.assertEqual(len(rcr.passing_measurements()), 2)
+        self.assertEqual(len(rcr.failing_measurements()), 10)
+
+    def test_top_n_failing(self):
+        """
+        Test that the top N failing measurements are returned
+        if no passing measurements are present
+        """
+        rcr = self._rcr_throughput_with_latency_constraint
+
+        for i in range(2, 6):
+            self._add_rcm_to_rcr(rcr, throughput_value=10 * i, latency_value=100 * i)
+
+        # Failing measurements are returned most to least throughput
+        failing_non_gpu_data = [
+            convert_non_gpu_metrics_to_data(
+                {"perf_throughput": 10 * i, "perf_latency_p99": 100 * i}
+            )
+            for i in range(5, 1, -1)
+        ]
+
+        top_n_measurements = rcr.top_n_measurements(3)
+
+        for i in range(3):
+            self.assertEqual(
+                top_n_measurements[i].non_gpu_data(), [failing_non_gpu_data[i]]
+            )
+
+    def test_top_n_failing_multi_model(self):
+        """
+        Test that the top N failing measurements are returned
+        if no passing measurements are present in a multi-model
+        configuration
+        """
+        rcr = self._rcr_throughput_with_latency_constraint_multi_model
+
+        for i in range(1, 6):
+            self._add_multi_model_rcm_to_rcr(
+                rcr, throughput_values=[5 * i, 20 * i], latency_values=[50 * i, 200 * i]
+            )
+
+        # Failing measurements are returned most to least throughput
+        failing_non_gpu_data = []
+        failing_non_gpu_data.append(
+            [
+                convert_non_gpu_metrics_to_data(
+                    {"perf_throughput": 5 * i, "perf_latency_p99": 50 * i}
+                )
+                for i in range(5, 1, -1)
+            ]
+        )
+
+        failing_non_gpu_data.append(
+            [
+                convert_non_gpu_metrics_to_data(
+                    {"perf_throughput": 20 * i, "perf_latency_p99": 200 * i}
+                )
+                for i in range(5, 1, -1)
+            ]
+        )
+
+        top_n_measurements = rcr.top_n_measurements(3)
+
+        for i in range(3):
+            self.assertEqual(
+                top_n_measurements[i].non_gpu_data(),
+                [failing_non_gpu_data[0][i], failing_non_gpu_data[1][i]],
+            )
+
+    def test_top_n_passing(self):
+        """
+        Test that the top N passing measurements are returned
+        """
+        rcr = self._rcr_throughput_with_latency_constraint
+
+        # 4 passing, 6 failing
+        for i in range(1, 11):
+            self._add_rcm_to_rcr(rcr, throughput_value=10 * i, latency_value=25 * i)
+
+        # Passing measurements are returned most to least throughput
+        passing_non_gpu_data = [
+            convert_non_gpu_metrics_to_data(
+                {"perf_throughput": 10 * i, "perf_latency_p99": 25 * i}
+            )
+            for i in range(4, 0, -1)
+        ]
+
+        top_n_measurements = rcr.top_n_measurements(3)
+
+        for i in range(3):
+            self.assertEqual(
+                top_n_measurements[i].non_gpu_data(), [passing_non_gpu_data[i]]
+            )
+
+    def test_top_n_passing_multi_model(self):
+        """
+        Test that the top N passing measurements are returned
+        in a multi-model configuration
+        """
+
+        rcr = self._rcr_throughput_with_latency_constraint_multi_model
+
+        # 4 passing, 6 failing
+        for i in range(1, 11):
+            self._add_rcm_to_rcr(rcr, throughput_value=10 * i, latency_value=25 * i)
+
+        # Passing measurements are returned most to least throughput
+        passing_non_gpu_data = [
+            convert_non_gpu_metrics_to_data(
+                {"perf_throughput": 10 * i, "perf_latency_p99": 25 * i}
+            )
+            for i in range(4, 0, -1)
+        ]
+
+        top_n_measurements = rcr.top_n_measurements(3)
+
+        for i in range(3):
+            self.assertEqual(
+                top_n_measurements[i].non_gpu_data(), [passing_non_gpu_data[i]]
+            )
+
+    def _construct_empty_rcr(self):
+        self.model_name = MagicMock()
+        self.run_config = MagicMock()
+
+        self.rcr_empty = RunConfigResult(
+            model_name=self.model_name,
+            run_config=self.run_config,
+            comparator=MagicMock(),
+            constraint_manager=MagicMock(),
+        )
+
+    def _construct_throughput_with_latency_constraint_rcr(self):
+        self._rcr_throughput_with_latency_constraint = RunConfigResult(
+            model_name=MagicMock(),
+            run_config=MagicMock(),
+            comparator=[{"perf_throughput": 1}],
+            constraint_manager=self.default_constraint_manager,
+        )
+
+    def _construct_throughput_with_latency_constraint_multi_model_rcr(self):
+        self._rcr_throughput_with_latency_constraint_multi_model = RunConfigResult(
+            model_name=MagicMock(),
+            run_config=MagicMock(),
+            comparator=[{"perf_throughput": 1}],
+            constraint_manager=self.default_constraint_manager,
+        )
+
+    def _add_rcm_to_rcr(self, rcr, throughput_value, latency_value):
+        rcr.add_run_config_measurement(
+            self._construct_single_model_rcm(throughput_value, latency_value)
+        )
+
+    def _add_multi_model_rcm_to_rcr(self, rcr, throughput_values, latency_values):
+        rcr.add_run_config_measurement(
+            self._construct_multi_model_rcm(throughput_values, latency_values)
+        )
+
+    def _construct_single_model_rcm(self, throughput_value, latency_value):
+        return construct_run_config_measurement(
+            model_name="modelA",
+            model_config_names=["modelA_config_0"],
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=MagicMock(),
+            non_gpu_metric_values=[
+                {"perf_throughput": throughput_value, "perf_latency_p99": latency_value}
+            ],
+        )
+
+    def _construct_multi_model_rcm(self, throughput_values, latency_values):
+        return construct_run_config_measurement(
+            model_name="modelA,modelB",
+            model_config_names=["modelA_config_0", "modelB_config_0"],
+            model_specific_pa_params=MagicMock(),
+            gpu_metric_values=MagicMock(),
+            non_gpu_metric_values=[
+                {
+                    "perf_throughput": throughput_values[0],
+                    "perf_latency_p99": latency_values[0],
+                },
+                {
+                    "perf_throughput": throughput_values[1],
+                    "perf_latency_p99": latency_values[1],
+                },
+            ],
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_run_config_result_comparator.py b/src/c++/perf_analyzer/tests/test_run_config_result_comparator.py
new file mode 100755
index 000000000..34a210603
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_run_config_result_comparator.py
@@ -0,0 +1,258 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from typing import List
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.result.run_config_result_comparator import RunConfigResultComparator
+
+from .common import test_result_collector as trc
+from .common.test_utils import construct_run_config_result
+
+
+class TestRunConfigResultComparatorMethods(trc.TestResultCollector):
+    def setUp(self):
+        self._initialize_metrics()
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_throughput_driven(self):
+        objective_spec = [{"perf_throughput": 2, "perf_latency_p99": 1}]
+        model_weights = [1]
+
+        self._check_run_config_result_comparison(
+            objective_spec=objective_spec,
+            model_weights=model_weights,
+            avg_gpu_metrics1=self.avg_gpu_metrics1,
+            avg_non_gpu_metrics1=self.avg_non_gpu_metrics1,
+            avg_gpu_metrics2=self.avg_gpu_metrics2,
+            avg_non_gpu_metrics2=self.avg_non_gpu_metrics2,
+            expected_result=False,
+        )
+
+    def test_throughput_driven_multi_model(self):
+        objective_spec = [
+            {"perf_throughput": 2, "perf_latency_p99": 1},
+            {"perf_throughput": 2, "perf_latency_p99": 1},
+        ]
+        model_weights = [1, 1]
+
+        self._check_run_config_result_comparison(
+            objective_spec=objective_spec,
+            model_weights=model_weights,
+            avg_gpu_metrics1=self.avg_gpu_metrics1,
+            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_multi1,
+            avg_gpu_metrics2=self.avg_gpu_metrics2,
+            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_multi2,
+            expected_result=False,
+            model_name="test_model",
+            model_config_names=["test_model_config_0", "test_model_config_1"],
+        )
+
+    def test_latency_driven(self):
+        objective_spec = [
+            {"perf_throughput": 1, "perf_latency_p99": 2},
+        ]
+        model_weights = [1]
+
+        self._check_run_config_result_comparison(
+            objective_spec=objective_spec,
+            model_weights=model_weights,
+            avg_gpu_metrics1=self.avg_gpu_metrics1,
+            avg_non_gpu_metrics1=self.avg_non_gpu_metrics1,
+            avg_gpu_metrics2=self.avg_gpu_metrics2,
+            avg_non_gpu_metrics2=self.avg_non_gpu_metrics2,
+            expected_result=True,
+        )
+
+    def test_latency_driven_multi(self):
+        objective_spec = [
+            {"perf_throughput": 1, "perf_latency_p99": 2},
+            {"perf_throughput": 1, "perf_latency_p99": 2},
+        ]
+        model_weights = [1, 1]
+
+        self._check_run_config_result_comparison(
+            objective_spec=objective_spec,
+            model_weights=model_weights,
+            avg_gpu_metrics1=self.avg_gpu_metrics1,
+            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_multi1,
+            avg_gpu_metrics2=self.avg_gpu_metrics2,
+            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_multi2,
+            expected_result=True,
+            model_name="test_model",
+            model_config_names=["test_model_config_0", "test_model_config_1"],
+        )
+
+    def test_equal_weight(self):
+        objective_spec = [{"perf_throughput": 1, "perf_latency_p99": 1}]
+        model_weights = [1]
+
+        self._check_run_config_result_comparison(
+            objective_spec=objective_spec,
+            model_weights=model_weights,
+            avg_gpu_metrics1=self.avg_gpu_metrics1,
+            avg_non_gpu_metrics1=self.avg_non_gpu_metrics1,
+            avg_gpu_metrics2=self.avg_gpu_metrics2,
+            avg_non_gpu_metrics2=self.avg_non_gpu_metrics2,
+            value_step2=2,
+            expected_result=False,
+        )
+
+    def test_equal_weight_multi(self):
+        objective_spec = [
+            {"perf_throughput": 1, "perf_latency_p99": 1},
+            {"perf_throughput": 1, "perf_latency_p99": 1},
+        ]
+        model_weights = [1, 1]
+
+        self._check_run_config_result_comparison(
+            objective_spec=objective_spec,
+            model_weights=model_weights,
+            avg_gpu_metrics1=self.avg_gpu_metrics1,
+            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_multi1,
+            avg_gpu_metrics2=self.avg_gpu_metrics2,
+            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_multi2,
+            value_step2=2,
+            expected_result=False,
+            model_name="test_model",
+            model_config_names=["test_model_config_0", "test_model_config_1"],
+        )
+
+    def test_unequal_weight_multi(self):
+        """
+        Tests that changing just the model weighting will result in a different
+        config being selected as better
+        """
+        objective_spec = [
+            {"perf_throughput": 1, "perf_latency_p99": 1},
+            {"perf_throughput": 1, "perf_latency_p99": 1},
+        ]
+
+        # With equal weighting - config 1 is better
+        model_weights = [1, 1]
+        self._check_run_config_result_comparison(
+            objective_spec=objective_spec,
+            model_weights=model_weights,
+            avg_gpu_metrics1=self.avg_gpu_metrics1,
+            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_weighted_multi1,
+            avg_gpu_metrics2=self.avg_gpu_metrics2,
+            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_weighted_multi2,
+            value_step2=2,
+            expected_result=False,
+            model_name="test_model",
+            model_config_names=["test_model_config_0", "test_model_config_1"],
+        )
+
+        # With an unequal weighting - config 0 is better
+        model_weights = [3, 1]
+        self._check_run_config_result_comparison(
+            objective_spec=objective_spec,
+            model_weights=model_weights,
+            avg_gpu_metrics1=self.avg_gpu_metrics1,
+            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_weighted_multi1,
+            avg_gpu_metrics2=self.avg_gpu_metrics2,
+            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_weighted_multi2,
+            value_step2=2,
+            expected_result=True,
+            model_name="test_model",
+            model_config_names=["test_model_config_0", "test_model_config_1"],
+        )
+
+    def _check_run_config_result_comparison(
+        self,
+        objective_spec,
+        model_weights: List[int],
+        avg_gpu_metrics1,
+        avg_gpu_metrics2,
+        avg_non_gpu_metrics1,
+        avg_non_gpu_metrics2,
+        value_step1=1,
+        value_step2=1,
+        expected_result=0,
+        model_name="test_model",
+        model_config_names=["test_model"],
+    ):
+        """
+        Helper function that takes all the data needed to
+        construct two RunConfigResults, constructs and runs a
+        comparator and checks that it produces the expected
+        value.
+        """
+
+        result_comparator = RunConfigResultComparator(
+            metric_objectives_list=objective_spec, model_weights=model_weights
+        )
+
+        result1 = construct_run_config_result(
+            avg_gpu_metric_values=avg_gpu_metrics1,
+            avg_non_gpu_metric_values_list=avg_non_gpu_metrics1,
+            comparator=result_comparator,
+            value_step=value_step1,
+            run_config=MagicMock(),
+            constraint_manager=MagicMock(),
+            model_name=model_name,
+            model_config_names=model_config_names,
+        )
+
+        result2 = construct_run_config_result(
+            avg_gpu_metric_values=avg_gpu_metrics2,
+            avg_non_gpu_metric_values_list=avg_non_gpu_metrics2,
+            comparator=result_comparator,
+            value_step=value_step2,
+            run_config=MagicMock(),
+            constraint_manager=MagicMock(),
+            model_name=model_name,
+            model_config_names=model_config_names,
+        )
+
+        self.assertEqual(
+            result_comparator.is_better_than(result1, result2), expected_result
+        )
+
+    def _initialize_metrics(self):
+        self.avg_gpu_metrics1 = {0: {"gpu_used_memory": 5000, "gpu_utilization": 50}}
+        self.avg_gpu_metrics2 = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+
+        self.avg_non_gpu_metrics1 = [{"perf_throughput": 100, "perf_latency_p99": 4000}]
+
+        self.avg_non_gpu_metrics2 = [{"perf_throughput": 200, "perf_latency_p99": 8000}]
+
+        self.avg_non_gpu_metrics_multi1 = [
+            {"perf_throughput": 100, "perf_latency_p99": 4000},
+            {"perf_throughput": 150, "perf_latency_p99": 6000},
+        ]
+
+        self.avg_non_gpu_metrics_multi2 = [
+            {"perf_throughput": 200, "perf_latency_p99": 8000},
+            {"perf_throughput": 250, "perf_latency_p99": 10000},
+        ]
+
+        self.avg_non_gpu_metrics_weighted_multi1 = [
+            {"perf_throughput": 1000, "perf_latency_p99": 50},
+            {"perf_throughput": 1000, "perf_latency_p99": 100},
+        ]
+
+        self.avg_non_gpu_metrics_weighted_multi2 = [
+            {"perf_throughput": 2000, "perf_latency_p99": 200},
+            {"perf_throughput": 500, "perf_latency_p99": 20},
+        ]
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_search_config.py b/src/c++/perf_analyzer/tests/test_search_config.py
new file mode 100755
index 000000000..b599ffb70
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_search_config.py
@@ -0,0 +1,88 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from model_analyzer.config.generate.search_config import SearchConfig
+from model_analyzer.config.generate.search_dimension import SearchDimension
+from model_analyzer.config.generate.search_dimensions import SearchDimensions
+
+from .common import test_result_collector as trc
+
+
+class TestSearchConfig(trc.TestResultCollector):
+    def test_basic(self):
+        sc = SearchConfig(SearchDimensions(), 0, 0)
+        self.assertEqual(0, sc.get_num_dimensions())
+
+    def test_config(self):
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        sc = SearchConfig(dimensions=dims, radius=4, min_initialized=2)
+
+        self.assertEqual(2, sc.get_num_dimensions())
+        self.assertEqual(4, sc.get_radius())
+        self.assertEqual(2, sc.get_min_initialized())
+
+        self.assertEqual("foo", sc.get_dimension(0).get_name())
+        self.assertEqual("bar", sc.get_dimension(1).get_name())
+
+        self.assertEqual(7, sc.get_dimension(0).get_value_at_idx(6))
+        self.assertEqual(64, sc.get_dimension(1).get_value_at_idx(6))
+
+    def test_get_min_indexes(self):
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR, 1, 10),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+        sc = SearchConfig(dims, 0, 0)
+
+        self.assertEqual([1, 0], sc.get_min_indexes())
+
+    def test_get_neighborhood_config(self):
+        """
+        Test that we can get a NeighborhoodConfig from a SearchConfig and properly override the radius
+        """
+        dims = SearchDimensions()
+        dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
+            ],
+        )
+
+        sc = SearchConfig(dimensions=dims, radius=4, min_initialized=2)
+
+        nc = sc.get_neighborhood_config(radius=5)
+        self.assertEqual(2, nc.get_num_dimensions())
+        self.assertEqual(5, nc.get_radius())
+        self.assertEqual(2, nc.get_min_initialized())
+
+        self.assertEqual("foo", nc.get_dimension(0).get_name())
+        self.assertEqual("bar", nc.get_dimension(1).get_name())
+
+        self.assertEqual(7, nc.get_dimension(0).get_value_at_idx(6))
+        self.assertEqual(64, nc.get_dimension(1).get_value_at_idx(6))
diff --git a/src/c++/perf_analyzer/tests/test_search_dimension.py b/src/c++/perf_analyzer/tests/test_search_dimension.py
new file mode 100755
index 000000000..1f51b2ea6
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_search_dimension.py
@@ -0,0 +1,53 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from model_analyzer.config.generate.search_dimension import SearchDimension
+
+from .common import test_result_collector as trc
+
+
+class TestSearchDimension(trc.TestResultCollector):
+    def test_linear(self):
+        sd = SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR, 0, 10)
+
+        self.assertEqual("foo", sd.get_name())
+        self.assertEqual(0, sd.get_min_idx())
+        self.assertEqual(10, sd.get_max_idx())
+        self.assertEqual(1, sd.get_value_at_idx(0))
+        self.assertEqual(2, sd.get_value_at_idx(1))
+        self.assertEqual(3, sd.get_value_at_idx(2))
+
+    def test_exponential(self):
+        sd = SearchDimension("foo", SearchDimension.DIMENSION_TYPE_EXPONENTIAL)
+        self.assertEqual(1, sd.get_value_at_idx(0))
+        self.assertEqual(2, sd.get_value_at_idx(1))
+        self.assertEqual(4, sd.get_value_at_idx(2))
+        self.assertEqual(8, sd.get_value_at_idx(3))
+
+    def test_out_of_bounds(self):
+        sd = SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR, 2, 10)
+
+        with self.assertRaises(IndexError):
+            sd.get_value_at_idx(1)
+
+        with self.assertRaises(IndexError):
+            sd.get_value_at_idx(11)
+
+    def test_no_max(self):
+        sd = SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR)
+
+        # Confirm no assert
+        sd.get_value_at_idx(100000000)
diff --git a/src/c++/perf_analyzer/tests/test_search_dimensions.py b/src/c++/perf_analyzer/tests/test_search_dimensions.py
new file mode 100755
index 000000000..e9a0e79ad
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_search_dimensions.py
@@ -0,0 +1,79 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from model_analyzer.config.generate.coordinate import Coordinate
+from model_analyzer.config.generate.search_dimension import SearchDimension
+from model_analyzer.config.generate.search_dimensions import SearchDimensions
+
+from .common import test_result_collector as trc
+
+
+class TestSearchDimensions(trc.TestResultCollector):
+    def setUp(self):
+        self._dims = SearchDimensions()
+        self._dims.add_dimensions(
+            0,
+            [
+                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
+                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_LINEAR),
+            ],
+        )
+
+        self._dims.add_dimensions(
+            1, [SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR)]
+        )
+
+    def test_list_behavior(self):
+        """
+        Confirm the following list-like behavior:
+            - get length like a list
+            - get from index like a list
+            - iterate over values
+            - raise IndexError on out of bounds index
+        """
+        dims = self._dims
+        self.assertEqual(len(dims), 3)
+
+        dim0 = dims[0]
+        dim1 = dims[1]
+        dim2 = dims[2]
+        self.assertEqual(dim0.get_name(), "foo")
+        self.assertEqual(dim1.get_name(), "bar")
+        self.assertEqual(dim2.get_name(), "foo")
+
+        count = 0
+        for dim in dims:
+            self.assertTrue(isinstance(dim, SearchDimension))
+            count += 1
+        self.assertEqual(count, 3)
+
+        with self.assertRaises(IndexError):
+            _ = dims[3]
+
+    def test_get_values(self):
+        """
+        Test get_values_for_coordinate() functionality
+        """
+        dims = self._dims
+        vals = dims.get_values_for_coordinate(Coordinate([0, 2, 4]))
+        expected_vals = {
+            0: {
+                "foo": 1,
+                "bar": 3,
+            },
+            1: {"foo": 5},
+        }
+        self.assertEqual(vals, expected_vals)
diff --git a/src/c++/perf_analyzer/tests/test_sorted_results.py b/src/c++/perf_analyzer/tests/test_sorted_results.py
new file mode 100755
index 000000000..1b8c76715
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_sorted_results.py
@@ -0,0 +1,306 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from random import sample
+from unittest.mock import MagicMock, patch
+
+import yaml
+
+from model_analyzer.result.run_config_result_comparator import RunConfigResultComparator
+from model_analyzer.result.sorted_results import SortedResults
+
+from .common import test_result_collector as trc
+from .common.test_utils import (
+    construct_constraint_manager,
+    construct_run_config,
+    construct_run_config_result,
+)
+
+
+class TestSortedResults(trc.TestResultCollector):
+    def setUp(self):
+        objective_spec = {"perf_throughput": 10, "perf_latency_p99": 5}
+        self.sorted_results = SortedResults()
+        self.result_comparator = RunConfigResultComparator(
+            metric_objectives_list=[objective_spec], model_weights=[1]
+        )
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_add_results(self):
+        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+        avg_non_gpu_metrics = {"perf_throughput": 100, "perf_latency_p99": 4000}
+
+        for _ in range(10):
+            self.sorted_results.add_result(
+                construct_run_config_result(
+                    avg_gpu_metric_values=avg_gpu_metrics,
+                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
+                    constraint_manager=MagicMock(),
+                    comparator=self.result_comparator,
+                )
+            )
+
+        results = self.sorted_results.results()
+        self.assertEqual(len(results), 10)
+
+    def test_add_results_same_config(self):
+        """
+        Test that additional results with the same model config variant
+        will be added to the same entry in the list
+        """
+        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+        avg_non_gpu_metrics = {"perf_throughput": 100, "perf_latency_p99": 4000}
+
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              test_model
+            """
+        )
+        for _ in range(10):
+            run_config = construct_run_config("modelA", "model_config_0", "key_A")
+            self.sorted_results.add_result(
+                construct_run_config_result(
+                    avg_gpu_metric_values=avg_gpu_metrics,
+                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
+                    comparator=self.result_comparator,
+                    constraint_manager=constraint_manager,
+                    run_config=run_config,
+                )
+            )
+
+        results = self.sorted_results.results()
+        self.assertEqual(len(results), 1)
+
+    def test_add_results_dynamic_sorting(self):
+        """
+        Tests the case where configs A & B where A is better than B
+        Then add a new measurement to B, which now makes B better than A
+        """
+        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+
+        run_config_A = construct_run_config("model", "model_config_A", "key_A")
+        run_config_B = construct_run_config("model", "model_config_B", "key_B")
+        run_config_list = [run_config_A, run_config_B]
+
+        constraint_manager = construct_constraint_manager(
+            """
+            profile_models:
+              model
+            """
+        )
+
+        for i in sample(range(2), 2):
+            avg_non_gpu_metrics = {
+                "perf_throughput": 100 - 10 * i,
+                "perf_latency_p99": 4000,
+            }
+            self.sorted_results.add_result(
+                construct_run_config_result(
+                    avg_gpu_metric_values=avg_gpu_metrics,
+                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
+                    comparator=self.result_comparator,
+                    model_name="model",
+                    model_config_names=["model_config_0"],
+                    constraint_manager=constraint_manager,
+                    run_config=run_config_list[i],
+                )
+            )
+
+        all_results = self.sorted_results.top_n_results(n=SortedResults.GET_ALL_RESULTS)
+
+        self.assertEqual(
+            all_results[0].run_config().model_variants_name(), "model_config_A"
+        )
+        self.assertEqual(
+            all_results[1].run_config().model_variants_name(), "model_config_B"
+        )
+
+        avg_non_gpu_metrics = {"perf_throughput": 200, "perf_latency_p99": 4000}
+
+        self.sorted_results.add_result(
+            construct_run_config_result(
+                avg_gpu_metric_values=avg_gpu_metrics,
+                avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
+                comparator=self.result_comparator,
+                model_name="model",
+                model_config_names=["model_config_0"],
+                constraint_manager=constraint_manager,
+                run_config=run_config_B,
+            )
+        )
+
+        all_results = self.sorted_results.top_n_results(n=SortedResults.GET_ALL_RESULTS)
+
+        self.assertEqual(
+            all_results[0].run_config().model_variants_name(), "model_config_B"
+        )
+        self.assertEqual(
+            all_results[1].run_config().model_variants_name(), "model_config_A"
+        )
+
+    def test_add_results_all_failing(self):
+        """
+        Test the case where we have only failing results
+        """
+        # Create constraint_manager for all model names
+        constraint_manager = construct_constraint_manager(
+            yaml.dump(
+                {
+                    "profile_models": {
+                        str(i): {"constraints": {"perf_throughput": {"min": 1000}}}
+                        for i in range(10)
+                    }
+                }
+            )
+        )
+
+        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+        for i in sample(range(10), 10):
+            model_name = str(i)
+            model_config_names = [f"{model_name}_config_0"]
+
+            avg_non_gpu_metrics = {
+                "perf_throughput": 100 + 10 * i,
+                "perf_latency_p99": 4000,
+            }
+            self.sorted_results.add_result(
+                construct_run_config_result(
+                    avg_gpu_metric_values=avg_gpu_metrics,
+                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
+                    comparator=self.result_comparator,
+                    constraint_manager=constraint_manager,
+                    model_config_names=model_config_names,
+                    model_name=model_name,
+                )
+            )
+
+        top_5_results = self.sorted_results.top_n_results(n=5)
+        self.assertEqual(top_5_results[0].model_name(), "9")
+        self.assertEqual(top_5_results[1].model_name(), "8")
+        self.assertEqual(top_5_results[2].model_name(), "7")
+        self.assertEqual(top_5_results[3].model_name(), "6")
+        self.assertEqual(top_5_results[4].model_name(), "5")
+
+    def test_add_results_failing_to_passing(self):
+        """
+        Test the case where we have only failing results
+        and then a measurement makes one of the results passing
+        """
+
+        # Create constraint_manager for all model names
+        constraint_manager = construct_constraint_manager(
+            yaml.dump(
+                {
+                    "profile_models": {
+                        str(i): {"constraints": {"perf_throughput": {"min": 1000}}}
+                        for i in sample(range(10), 10)
+                    }
+                }
+            )
+        )
+
+        # Create 10 failing results
+        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+        for i in sample(range(10), 10):
+            model_name = str(i)
+            model_config_names = [f"{model_name}_config_0"]
+
+            avg_non_gpu_metrics = {
+                "perf_throughput": 100 + 10 * i,
+                "perf_latency_p99": 4000,
+            }
+            self.sorted_results.add_result(
+                construct_run_config_result(
+                    avg_gpu_metric_values=avg_gpu_metrics,
+                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
+                    comparator=self.result_comparator,
+                    constraint_manager=constraint_manager,
+                    model_config_names=model_config_names,
+                    model_name=model_name,
+                )
+            )
+
+        # Now add a measurement to the last result so that it now passes the constraint
+        avg_non_gpu_metrics = {"perf_throughput": 2000, "perf_latency_p99": 4000}
+
+        model_name = "9"
+        model_config_names = [f"{model_name}_config_0"]
+        self.sorted_results.add_result(
+            construct_run_config_result(
+                avg_gpu_metric_values=avg_gpu_metrics,
+                avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
+                comparator=self.result_comparator,
+                constraint_manager=constraint_manager,
+                model_config_names=model_config_names,
+                model_name=model_name,
+            )
+        )
+
+        top_5_results = self.sorted_results.top_n_results(n=5)
+        self.assertEqual(len(top_5_results), 1)
+        self.assertEqual(top_5_results[0].model_name(), "9")
+
+    def test_top_n_results(self):
+        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
+        constraint_manager = construct_constraint_manager(
+            yaml.dump({"profile_models": {str(i): {} for i in range(10)}})
+        )
+
+        for i in sample(range(10), 10):
+            model_name = str(i)
+            model_config_names = [f"{model_name}_config_0"]
+
+            avg_non_gpu_metrics = {
+                "perf_throughput": 100 + 10 * i,
+                "perf_latency_p99": 4000,
+            }
+            self.sorted_results.add_result(
+                construct_run_config_result(
+                    avg_gpu_metric_values=avg_gpu_metrics,
+                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
+                    comparator=self.result_comparator,
+                    model_name=model_name,
+                    model_config_names=model_config_names,
+                    constraint_manager=constraint_manager,
+                )
+            )
+
+        top_5_results = self.sorted_results.top_n_results(n=5)
+        self.assertEqual(top_5_results[0].model_name(), "9")
+        self.assertEqual(top_5_results[1].model_name(), "8")
+        self.assertEqual(top_5_results[2].model_name(), "7")
+        self.assertEqual(top_5_results[3].model_name(), "6")
+        self.assertEqual(top_5_results[4].model_name(), "5")
+
+        all_results = self.sorted_results.top_n_results(n=SortedResults.GET_ALL_RESULTS)
+        self.assertEqual(all_results[0].model_name(), "9")
+        self.assertEqual(all_results[1].model_name(), "8")
+        self.assertEqual(all_results[2].model_name(), "7")
+        self.assertEqual(all_results[3].model_name(), "6")
+        self.assertEqual(all_results[4].model_name(), "5")
+        self.assertEqual(all_results[5].model_name(), "4")
+        self.assertEqual(all_results[6].model_name(), "3")
+        self.assertEqual(all_results[7].model_name(), "2")
+        self.assertEqual(all_results[8].model_name(), "1")
+        self.assertEqual(all_results[9].model_name(), "0")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_state_manager.py b/src/c++/perf_analyzer/tests/test_state_manager.py
new file mode 100755
index 000000000..2a5898210
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_state_manager.py
@@ -0,0 +1,148 @@
+#!/usr/bin/env python3
+
+# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
+
+from .common import test_result_collector as trc
+from .common.test_utils import evaluate_mock_config
+from .mocks.mock_glob import MockGlobMethods
+from .mocks.mock_io import MockIOMethods
+from .mocks.mock_json import MockJSONMethods
+from .mocks.mock_os import MockOSMethods
+
+
+class TestAnalyzerStateManagerMethods(trc.TestResultCollector):
+    @patch(
+        "model_analyzer.config.input.config_command_profile.ConfigCommandProfile._preprocess_and_verify_arguments",
+        MagicMock(),
+    )
+    def setUp(self):
+        args = [
+            "model-analyzer",
+            "profile",
+            "--model-repository",
+            "cli_repository",
+            "-f",
+            "path-to-config-file",
+            "--profile-models",
+            "test_model",
+        ]
+        yaml_str = """
+            export_path: /test_export_path/
+        """
+
+        # start mocks
+        self.mock_io = MockIOMethods(
+            mock_paths=["model_analyzer.state.analyzer_state_manager"]
+        )
+        self.mock_json = MockJSONMethods()
+        self.mock_os = MockOSMethods(
+            mock_paths=[
+                "model_analyzer.state.analyzer_state_manager",
+                "model_analyzer.config.input.config_utils",
+            ]
+        )
+        self.mock_glob = MockGlobMethods()
+
+        self.mock_io.start()
+        self.mock_json.start()
+        self.mock_os.start()
+        self.mock_glob.start()
+
+        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
+
+        # state manager
+        self.state_manager = AnalyzerStateManager(config=config, server=None)
+
+    def tearDown(self):
+        patch.stopall()
+
+    def test_set_get_state_variables(self):
+        self.mock_os.set_os_path_exists_return_value(False)
+        self.state_manager.load_checkpoint(checkpoint_required=False)
+
+        vars = [f"test_var{j}" for j in range(10)]
+        for i, name in enumerate(vars):
+            self.state_manager.set_state_variable(name, i)
+
+        for i, name in enumerate(vars):
+            self.assertEqual(self.state_manager.get_state_variable(name), i)
+
+        for i, name in enumerate(vars):
+            self.state_manager.set_state_variable(name, 9 - i)
+
+        for i, name in enumerate(vars):
+            self.assertEqual(self.state_manager.get_state_variable(name), 9 - i)
+
+    def test_load_checkpoint(self):
+        # Load checkpoint without ckpt files
+        self.mock_os.set_os_path_exists_return_value(False)
+        self.state_manager.load_checkpoint(checkpoint_required=False)
+        self.assertTrue(self.state_manager.starting_fresh_run())
+
+        # Load checkpoint files with ckpt files
+        self.mock_os.set_os_path_exists_return_value(True)
+        self.mock_os.set_os_path_join_return_value("0.ckpt")
+        self.state_manager.load_checkpoint(checkpoint_required=False)
+        self.assertFalse(self.state_manager.starting_fresh_run())
+
+        # Load checkpoint throws error
+        self.mock_json.set_json_load_side_effect(EOFError)
+        with self.assertRaises(
+            TritonModelAnalyzerException,
+            msg="Checkpoint file 0.ckpt is"
+            " empty or corrupted. Remove it from checkpoint"
+            " directory.",
+        ):
+            self.mock_os.set_os_path_exists_return_value(True)
+            self.mock_os.set_os_path_join_return_value("0.ckpt")
+            self.state_manager.load_checkpoint(checkpoint_required=False)
+            self.assertFalse(self.state_manager.starting_fresh_run())
+
+    def test_latest_checkpoint(self):
+        # No checkpoints
+        self.mock_glob.set_glob_return_value([])
+        self.assertEqual(self.state_manager._latest_checkpoint(), -1)
+
+        # single checkpoint file
+        for i in range(5):
+            self.mock_glob.set_glob_return_value([f"{i}.ckpt"])
+            self.assertEqual(self.state_manager._latest_checkpoint(), i)
+
+        # Multiple checkpoint files consecutive, sorted
+        self.mock_glob.set_glob_return_value([f"{i}.ckpt" for i in range(5)])
+        self.assertEqual(self.state_manager._latest_checkpoint(), 4)
+
+        # Multiple checkpoint files consecutive, unsorted
+        self.mock_glob.set_glob_return_value([f"{i}.ckpt" for i in range(5, 1, -1)])
+        self.assertEqual(self.state_manager._latest_checkpoint(), 5)
+
+        # Multiple files nonconsecutive unsorted
+        self.mock_glob.set_glob_return_value([f"{i}.ckpt" for i in [1, 3, 5, 2, 0, 4]])
+        self.assertEqual(self.state_manager._latest_checkpoint(), 5)
+
+        # Malformed checkpoint filename
+        self.mock_glob.set_glob_return_value(["XYZ.ckpt"])
+        with self.assertRaises(TritonModelAnalyzerException):
+            self.state_manager._latest_checkpoint()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_triton_client.py b/src/c++/perf_analyzer/tests/test_triton_client.py
new file mode 100755
index 000000000..f63623e5e
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_triton_client.py
@@ -0,0 +1,214 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import tempfile
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.device.gpu_device import GPUDevice
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.triton.client.client_factory import TritonClientFactory
+from model_analyzer.triton.server.server_config import TritonServerConfig
+from model_analyzer.triton.server.server_factory import TritonServerFactory
+
+from .common import test_result_collector as trc
+from .mocks.mock_client import MockTritonClientMethods
+from .mocks.mock_server_docker import MockServerDockerMethods
+
+# Test parameters
+MODEL_REPOSITORY_PATH = "test_repo"
+TRITON_IMAGE = "test_image"
+CONFIG_TEST_ARG = "url"
+GRPC_URL = "test_grpc_url"
+HTTP_URL = "test_http_url"
+TEST_MODEL_NAME = "test_model"
+
+
+class TestTritonClientMethods(trc.TestResultCollector):
+    def setUp(self):
+        # GPUs
+        gpus = [GPUDevice("TEST_DEVICE_NAME", 0, "TEST_PCI_BUS_ID", "TEST_UUID")]
+
+        # Mocks
+        self.server_docker_mock = MockServerDockerMethods()
+        self.tritonclient_mock = MockTritonClientMethods()
+        self.server_docker_mock.start()
+        self.tritonclient_mock.start()
+
+        # Create server config
+        self.server_config = TritonServerConfig()
+        self.server_config["model-repository"] = MODEL_REPOSITORY_PATH
+        self.server_config["model-control-mode"] = "explicit"
+
+        # Set CUDA_VISIBLE_DEVICES
+        os.environ["CUDA_VISIBLE_DEVICES"] = "0"
+
+        # Create and start the server
+        self.server = TritonServerFactory.create_server_docker(
+            image=TRITON_IMAGE, config=self.server_config, gpus=gpus
+        )
+
+    def test_create_client(self):
+        # Create GRPC client
+        TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
+        self.tritonclient_mock.assert_created_grpc_client_with_args(url=GRPC_URL)
+
+        # Create HTTP client
+        TritonClientFactory.create_http_client(server_url=HTTP_URL)
+        self.tritonclient_mock.assert_created_http_client_with_args(url=HTTP_URL)
+
+    def test_wait_for_server_ready(self):
+        # For reuse
+        def _test_with_client(self, client):
+            with self.assertRaises(
+                TritonModelAnalyzerException,
+                msg="Expected Exception trying" " wait for server ready",
+            ):
+                self.tritonclient_mock.raise_exception_on_wait_for_server_ready()
+                client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
+            self.tritonclient_mock.reset()
+
+            with self.assertRaises(
+                TritonModelAnalyzerException,
+                msg="Expected Exception on" " server not ready",
+            ):
+                self.tritonclient_mock.set_server_not_ready()
+                client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
+
+            self.tritonclient_mock.reset()
+            client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
+
+        # HTTP client
+        client = TritonClientFactory.create_http_client(server_url=HTTP_URL)
+        self.tritonclient_mock.assert_created_http_client_with_args(HTTP_URL)
+        _test_with_client(self, client)
+        self.tritonclient_mock.assert_http_client_waited_for_server_ready()
+
+        # GRPC client
+        client = TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
+        self.tritonclient_mock.assert_created_grpc_client_with_args(GRPC_URL)
+        _test_with_client(self, client)
+        self.tritonclient_mock.assert_grpc_client_waited_for_server_ready()
+
+    def test_wait_for_server_ready_with_invalid_argument(self):
+        """
+        Tests that we detect when an invalid argument is passed to the server
+        """
+        log_file = tempfile.NamedTemporaryFile()
+        log_file.write(b"Unexpected argument: UNKNOWN_CMD")
+
+        client = TritonClientFactory.create_http_client(server_url=HTTP_URL)
+        self.tritonclient_mock.raise_exception_on_wait_for_server_ready()
+
+        with self.assertRaises(TritonModelAnalyzerException):
+            client.wait_for_server_ready(
+                num_retries=1, sleep_time=0.1, log_file=log_file
+            )
+
+    def test_wait_for_model_ready(self):
+        # For reuse
+        def _test_with_client(self, client):
+            client.wait_for_model_ready(
+                model_name=TEST_MODEL_NAME, num_retries=1, sleep_time=0.1
+            )
+            self.tritonclient_mock.reset()
+
+            self.tritonclient_mock.set_model_not_ready()
+            self.assertTrue(
+                client.wait_for_model_ready(
+                    model_name=TEST_MODEL_NAME, num_retries=2, sleep_time=0.1
+                ),
+                -1,
+            )
+            self.tritonclient_mock.reset()
+            client.wait_for_model_ready(
+                model_name=TEST_MODEL_NAME, num_retries=1, sleep_time=0.1
+            )
+
+        # HTTP client
+        client = TritonClientFactory.create_http_client(server_url=HTTP_URL)
+        self.tritonclient_mock.assert_created_http_client_with_args(HTTP_URL)
+        _test_with_client(self, client)
+        self.tritonclient_mock.assert_http_client_waited_for_model_ready(
+            model_name=TEST_MODEL_NAME
+        )
+
+        # GRPC client
+        client = TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
+        self.tritonclient_mock.assert_created_grpc_client_with_args(GRPC_URL)
+        _test_with_client(self, client)
+        self.tritonclient_mock.assert_grpc_client_waited_for_model_ready(
+            model_name=TEST_MODEL_NAME
+        )
+
+    def test_load_unload_model(self):
+        # Create client
+        client = TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
+        self.tritonclient_mock.assert_created_grpc_client_with_args(GRPC_URL)
+
+        # Start the server and wait till it is ready
+        self.server.start()
+        client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
+
+        # Try to load a dummy model and expect error
+        self.tritonclient_mock.raise_exception_on_load()
+        self.assertTrue(client.load_model("dummy"), -1)
+
+        self.tritonclient_mock.reset()
+
+        # Load the test model
+        client.load_model(TEST_MODEL_NAME)
+        client.wait_for_model_ready(TEST_MODEL_NAME, num_retries=1, sleep_time=0.1)
+
+        # Try to unload a dummy model and expect error
+        self.tritonclient_mock.raise_exception_on_unload()
+        self.assertTrue(client.unload_model("dummy"), -1)
+
+        self.tritonclient_mock.reset()
+
+        # Unload the test model
+        client.unload_model(TEST_MODEL_NAME)
+
+        self.server.stop()
+
+    def test_get_model_config(self):
+        # Create client
+        client = TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
+        self.tritonclient_mock.assert_created_grpc_client_with_args(GRPC_URL)
+
+        # Start the server and wait till it is ready
+        self.server.start()
+        client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
+
+        # Set model config, and try to get it
+        test_model_config_dict = {"config": "test_config"}
+        self.tritonclient_mock.set_model_config(test_model_config_dict)
+        self.assertEqual(
+            client.get_model_config(TEST_MODEL_NAME, num_retries=1), "test_config"
+        )
+
+    def tearDown(self):
+        # In case test raises exception
+        if self.server is not None:
+            self.server.stop()
+        self.server_docker_mock.stop()
+        self.tritonclient_mock.stop()
+        patch.stopall()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_triton_server.py b/src/c++/perf_analyzer/tests/test_triton_server.py
new file mode 100755
index 000000000..b083b7567
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_triton_server.py
@@ -0,0 +1,361 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import patch
+
+from model_analyzer.device.gpu_device import GPUDevice
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+from model_analyzer.triton.server.server_config import TritonServerConfig
+from model_analyzer.triton.server.server_factory import TritonServerFactory
+
+from .common import test_result_collector as trc
+from .mocks.mock_gpu_device_factory import MockGPUDeviceFactory
+from .mocks.mock_os import MockOSMethods
+from .mocks.mock_server_docker import MockServerDockerMethods
+from .mocks.mock_server_local import MockServerLocalMethods
+
+# Test parameters
+MODEL_REPOSITORY_PATH = "test_repo"
+TRITON_LOCAL_BIN_PATH = "test_bin_path/tritonserver"
+TRITON_DOCKER_BIN_PATH = "tritonserver"
+TRITON_IMAGE = "test_image"
+CONFIG_TEST_ARG = "exit-on-error"
+CLI_TO_STRING_TEST_ARGS = {
+    "allow-grpc": True,
+    "min-supported-compute-capability": 7.5,
+    "metrics-port": 8000,
+    "model-repository": MODEL_REPOSITORY_PATH,
+}
+
+
+class TestTritonServerMethods(trc.TestResultCollector):
+    def setUp(self):
+        # GPUs for this test
+        self._sys_gpus = [
+            GPUDevice("TEST_DEVICE_NAME", 0, "TEST_BUS_ID0", "TEST_UUID0"),
+            GPUDevice("TEST_DEVICE_NAME", 1, "TEST_BUS_ID1", "TEST_UUID1"),
+        ]
+
+        # Mock
+        self.gpu_device_factory_mock = MockGPUDeviceFactory(self._sys_gpus)
+        self.server_docker_mock = MockServerDockerMethods()
+        self.server_local_mock = MockServerLocalMethods()
+        self.os_mock = MockOSMethods(
+            mock_paths=[
+                "model_analyzer.triton.server.server_local",
+                "tests.mocks.mock_server_local",
+            ]
+        )
+        self.gpu_device_factory_mock.start()
+        self.server_docker_mock.start()
+        self.server_local_mock.start()
+        self.os_mock.start()
+
+        # server setup
+        self.server = None
+
+    def test_server_config(self):
+        # Create a TritonServerConfig
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+
+        # Check config initializations
+        self.assertIsNone(
+            server_config[CONFIG_TEST_ARG],
+            msg="Server config had unexpected initial" f"value for {CONFIG_TEST_ARG}",
+        )
+        # Set value
+        server_config[CONFIG_TEST_ARG] = True
+
+        # Test get again
+        self.assertTrue(
+            server_config[CONFIG_TEST_ARG], msg=f"{CONFIG_TEST_ARG} was not set"
+        )
+
+        # Try to set an unsupported config argument, expect failure
+        with self.assertRaises(
+            TritonModelAnalyzerException,
+            msg="Expected exception on trying to set"
+            "unsupported argument in Triton server"
+            "config",
+        ):
+            server_config["dummy"] = 1
+
+        # Reset test arg
+        server_config[CONFIG_TEST_ARG] = None
+
+        # Finally set a couple of args and then check the cli string
+        for arg, value in CLI_TO_STRING_TEST_ARGS.items():
+            server_config[arg] = value
+
+        cli_string = server_config.to_cli_string()
+        for argstring in cli_string.split():
+            # Parse the created string
+            arg, value = argstring.split("=")
+            arg = arg[2:]
+
+            # Make sure each parsed arg was in test dict
+            self.assertIn(
+                arg,
+                CLI_TO_STRING_TEST_ARGS,
+                msg=f"CLI string contained unknown argument: {arg}",
+            )
+
+            # Make sure parsed value is the one from dict, check type too
+            test_value = CLI_TO_STRING_TEST_ARGS[arg]
+            self.assertEqual(
+                test_value,
+                type(test_value)(value),
+                msg=f"CLI string contained unknown value: {value}",
+            )
+
+    def _test_create_server(self, gpus):
+        # Create a TritonServerConfig
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+
+        # Run for both types of environments
+        self.server = TritonServerFactory.create_server_docker(
+            image=TRITON_IMAGE, config=server_config, gpus=gpus
+        )
+
+        self.server = TritonServerFactory.create_server_local(
+            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
+        )
+
+        # Try to create a server without specifying model repository and expect
+        # error
+        server_config["model-repository"] = None
+        with self.assertRaises(
+            AssertionError,
+            msg="Expected AssertionError for trying to create"
+            "server without specifying model repository.",
+        ):
+            self.server = TritonServerFactory.create_server_docker(
+                image=TRITON_IMAGE, config=server_config, gpus=gpus
+            )
+        with self.assertRaises(
+            AssertionError,
+            msg="Expected AssertionError for trying to create"
+            "server without specifying model repository.",
+        ):
+            self.server = TritonServerFactory.create_server_local(
+                path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
+            )
+
+    def test_triton_server_ssl_options(self):
+        server_config = TritonServerConfig()
+
+        triton_server_flags = {
+            "grpc-use-ssl": "1",
+            "grpc-use-ssl-mutual": "1",
+            "grpc-server-cert": "a",
+            "grpc-server-key": "b",
+            "grpc-root-cert": "c",
+        }
+        server_config.update_config(triton_server_flags)
+
+        expected_cli_str = (
+            f"--grpc-use-ssl=1 --grpc-use-ssl-mutual=1 "
+            f"--grpc-server-cert=a --grpc-server-key=b --grpc-root-cert=c"
+        )
+        self.assertEqual(server_config.to_cli_string(), expected_cli_str)
+
+    def test_create_server_no_gpu(self):
+        self._test_create_server(gpus=[])
+
+    def test_create_server_select_gpu(self):
+        self._test_create_server(gpus=self._sys_gpus[:1])
+
+    def _test_start_stop_gpus(self, gpus):
+        device_requests = [device.device_id() for device in gpus]
+
+        # Create a TritonServerConfig
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+
+        # Create server in docker, start , wait, and stop
+        self.server = TritonServerFactory.create_server_docker(
+            image=TRITON_IMAGE, config=server_config, gpus=gpus
+        )
+
+        # Start server check that mocked api is called
+        self.server.start()
+        self.server_docker_mock.assert_server_process_start_called_with(
+            f"{TRITON_DOCKER_BIN_PATH} {server_config.to_cli_string()}",
+            MODEL_REPOSITORY_PATH,
+            TRITON_IMAGE,
+            device_requests,
+            gpus,
+            8000,
+            8001,
+            8002,
+        )
+
+        self.server_docker_mock.raise_exception_on_container_run()
+        with self.assertRaises(TritonModelAnalyzerException):
+            self.server.start()
+        self.server_docker_mock.stop_raise_exception_on_container_run()
+
+        # Stop container and check api calls
+        self.server.stop()
+        self.server_docker_mock.assert_server_process_terminate_called()
+
+        # Create local server which runs triton as a subprocess
+        self.server = TritonServerFactory.create_server_local(
+            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
+        )
+
+        # Check that API functions are called
+        self.server.start()
+
+        self.server_local_mock.assert_server_process_start_called_with(
+            cmd=[TRITON_LOCAL_BIN_PATH, "--model-repository", MODEL_REPOSITORY_PATH],
+            gpus=gpus,
+            stdout=self.server._log_file,
+        )
+
+        self.server.stop()
+        self.server_local_mock.assert_server_process_terminate_called()
+
+    def test_start_stop_gpus_no_gpu(self):
+        self._test_start_stop_gpus(gpus=[])
+
+    def test_start_stop_gpus_select_gpu(self):
+        self._test_start_stop_gpus(gpus=self._sys_gpus[:1])
+
+    def start_stop_docker_args(self):
+        device_requests, gpu_uuids = self._find_correct_gpu_settings(self._sys_gpus)
+
+        # Create a TritonServerConfig
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+
+        # Create mounts and labels
+        mounts = ["/host/path:/dest/path:ro", "/another/host/path:/some/dest/path:rw"]
+        labels = {"RUNNER_ID": "TEST_RUNNER_ID"}
+
+        environment = {"VARIABLE": "VALUE"}
+        # Create server in docker, start , wait, and stop
+        self.server = TritonServerFactory.create_server_docker(
+            image=TRITON_IMAGE,
+            config=server_config,
+            gpus=self._sys_gpus,
+            mounts=mounts,
+            labels=labels,
+        )
+
+        # Start server check that mocked api is called
+        self.server.start(env=environment)
+        self.server_docker_mock.assert_server_process_start_called_with(
+            f"{TRITON_DOCKER_BIN_PATH} {server_config.to_cli_string()}",
+            MODEL_REPOSITORY_PATH,
+            TRITON_IMAGE,
+            device_requests,
+            gpu_uuids,
+            8000,
+            8001,
+            8002,
+            mounts,
+            labels,
+        )
+
+        # Stop container and check api calls
+        self.server.stop()
+        self.server_docker_mock.assert_server_process_terminate_called()
+
+    def _test_get_logs(self, gpus):
+        # Create a TritonServerConfig
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+
+        # Check docker server logs
+        self.server = TritonServerFactory.create_server_docker(
+            image=TRITON_IMAGE, config=server_config, gpus=gpus
+        )
+        self.server.start()
+        self.server.stop()
+        self.server_docker_mock.assert_server_process_terminate_called()
+        self.assertEqual(self.server.logs(), "Triton Server Test Log")
+
+        # Create local server logs
+        self.server = TritonServerFactory.create_server_local(
+            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
+        )
+        self.server.start()
+        self.server.stop()
+        self.server_local_mock.assert_server_process_terminate_called()
+        self.assertEqual(self.server.logs(), "Triton Server Test Log")
+
+    def _test_cpu_stats(self, gpus):
+        device_requests = [device.device_id() for device in gpus]
+
+        # Create a TritonServerConfig
+        server_config = TritonServerConfig()
+        server_config["model-repository"] = MODEL_REPOSITORY_PATH
+
+        # Test local server cpu_stats
+        self.server = TritonServerFactory.create_server_local(
+            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
+        )
+        self.server.start()
+        _, _ = self.server.cpu_stats()
+        self.server_local_mock.assert_cpu_stats_called()
+        self.server.stop()
+
+        # Test docker server cpu stats
+        self.server = TritonServerFactory.create_server_docker(
+            image=TRITON_IMAGE, config=server_config, gpus=gpus
+        )
+        self.server.start()
+
+        # The following needs to be called as it resets exec_run return value
+        self.server_docker_mock.assert_server_process_start_called_with(
+            f"{TRITON_DOCKER_BIN_PATH} {server_config.to_cli_string()}",
+            MODEL_REPOSITORY_PATH,
+            TRITON_IMAGE,
+            device_requests,
+            gpus,
+            8000,
+            8001,
+            8002,
+        )
+        _, _ = self.server.cpu_stats()
+        self.server_docker_mock.assert_cpu_stats_called()
+        self.server.stop()
+
+    def test_cpu_stats_no_gpu(self):
+        self._test_cpu_stats(gpus=[])
+
+    def test_cpu_stats_select_gpu(self):
+        self._test_cpu_stats(gpus=self._sys_gpus[:1])
+
+    def tearDown(self):
+        # In case test raises exception
+        if self.server is not None:
+            self.server.stop()
+
+        # Stop mocking
+        self.gpu_device_factory_mock.stop()
+        self.server_docker_mock.stop()
+        self.server_local_mock.stop()
+        self.os_mock.stop()
+        patch.stopall()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_triton_server_factory.py b/src/c++/perf_analyzer/tests/test_triton_server_factory.py
new file mode 100755
index 000000000..2448e89c9
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_triton_server_factory.py
@@ -0,0 +1,126 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+from unittest.mock import MagicMock, patch
+
+from model_analyzer.config.input.config_command_profile import ConfigCommandProfile
+from model_analyzer.triton.server.server_factory import TritonServerFactory
+
+from .common import test_result_collector as trc
+from .mocks.mock_os import MockOSMethods
+
+
+class TestTritonServerFactory(trc.TestResultCollector):
+    def setUp(self):
+        # Mock path validation
+        self.mock_os = MockOSMethods(
+            mock_paths=[
+                "model_analyzer.triton.server.server_factory",
+                "model_analyzer.config.input.config_utils",
+            ]
+        )
+        self.mock_os.start()
+
+    def _test_get_server_handle_helper(
+        self, launch_mode, expect_local, expect_docker, expect_config
+    ):
+        """
+        Test get_server_handle() calls the correct Triton Server create function with the
+        correct Triton Server Config options
+
+        Parameters:
+        ----------
+        launch_mode: str
+            which launch mode to use
+        expect_local: bool
+            If true, assert that create_server_local() is called
+        expect_docker: bool
+            If true, assert that create_server_docker() is called
+        expect_config: bool
+            If true, verify that certain Triton Server args are passed to the create function as expected
+            If false, verify that those args are NOT passed
+        """
+
+        config = ConfigCommandProfile()
+        config.model_repository = "/fake_model_repository"
+        config.triton_launch_mode = launch_mode
+        config.triton_http_endpoint = "fake_address:2345"
+        config.triton_grpc_endpoint = "fake_address:4567"
+        config.monitoring_interval = 0.5
+
+        expected_http_port = "2345"
+        expected_grpc_port = "4567"
+        # Convert seconds to ms
+        expected_metrics_interval_ms = int(config.monitoring_interval * 1000)
+
+        with patch(
+            "model_analyzer.triton.server.server_factory.TritonServerFactory.create_server_local"
+        ) as mocked_local, patch(
+            "model_analyzer.triton.server.server_factory.TritonServerFactory.create_server_docker"
+        ) as mocked_docker, patch(
+            "model_analyzer.triton.server.server_factory.TritonServerFactory._validate_triton_install_path"
+        ), patch(
+            "model_analyzer.triton.server.server_factory.TritonServerFactory._validate_triton_server_path"
+        ):
+            _ = TritonServerFactory.get_server_handle(config, MagicMock(), False)
+            self.assertEqual(mocked_local.call_count, expect_local)
+            self.assertEqual(mocked_docker.call_count, expect_docker)
+
+            if expect_local:
+                _, kwargs = mocked_local.call_args
+            else:
+                _, kwargs = mocked_docker.call_args
+            triton_config = kwargs["config"]
+
+            if expect_config:
+                self.assertEqual(triton_config["http-port"], expected_http_port)
+                self.assertEqual(triton_config["grpc-port"], expected_grpc_port)
+                self.assertEqual(
+                    triton_config["metrics-interval-ms"], expected_metrics_interval_ms
+                )
+
+            else:
+                self.assertEqual(triton_config["http-port"], None)
+                self.assertEqual(triton_config["grpc-port"], None)
+                self.assertEqual(triton_config["metrics-interval-ms"], None)
+
+    def test_get_server_handle_remote(self):
+        self._test_get_server_handle_helper(
+            launch_mode="remote", expect_local=1, expect_docker=0, expect_config=False
+        )
+
+    def test_get_server_handle_c_api(self):
+        self._test_get_server_handle_helper(
+            launch_mode="c_api", expect_local=1, expect_docker=0, expect_config=False
+        )
+
+    def test_get_server_handle_local(self):
+        self._test_get_server_handle_helper(
+            launch_mode="local", expect_local=1, expect_docker=0, expect_config=True
+        )
+
+    def test_get_server_handle_docker(self):
+        self._test_get_server_handle_helper(
+            launch_mode="docker", expect_local=0, expect_docker=1, expect_config=True
+        )
+
+    def tearDown(self):
+        patch.stopall()
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_yaml_options.py b/src/c++/perf_analyzer/tests/test_yaml_options.py
new file mode 100755
index 000000000..1c7ce9a4d
--- /dev/null
+++ b/src/c++/perf_analyzer/tests/test_yaml_options.py
@@ -0,0 +1,117 @@
+#!/usr/bin/env python3
+
+# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import yaml
+
+from model_analyzer.config.input.yaml_config_validator import YamlConfigValidator
+from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
+
+from .common import test_result_collector as trc
+
+
+class TestYamlOptions(trc.TestResultCollector):
+    def test_correct_option(self):
+        correct_option = "client_max_retries"
+        YamlConfigValidator._create_valid_option_set()
+        self.assertTrue(YamlConfigValidator._is_valid_option(correct_option))
+
+    def test_misspelled_option(self):
+        misspelled_option = "profile_model"
+        YamlConfigValidator._create_valid_option_set()
+        self.assertFalse(YamlConfigValidator._is_valid_option(misspelled_option))
+
+    def test_using_hyphens_not_underscores(self):
+        hyphen_option = "triton-server-flags"
+        YamlConfigValidator._create_valid_option_set()
+        self.assertFalse(YamlConfigValidator._is_valid_option(hyphen_option))
+
+    def test_multiple_options(self):
+        """
+        Tests multiple options.
+        The following are incorrect:
+            "profile_model",
+            "triton-server-flags",
+            "DURATION_seconds",
+            ""
+        """
+        options = {
+            "client_max_retries",
+            "profile_model",
+            "triton-server-flags",
+            "DURATION_seconds",
+            "",
+        }
+        count = 0
+        YamlConfigValidator._create_valid_option_set()
+        for entry in options:
+            if not YamlConfigValidator._is_valid_option(entry):
+                count += 1
+        self.assertEqual(
+            count,
+            4,
+            f"{count} incorrect yaml options are present in the yaml configuration file",
+        )
+
+    def test_valid_yaml_file(self):
+        # yapf: disable
+        yaml_str = ("""
+            run_config_search_max_instance_count: 16
+            run_config_search_disable: True
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        max_batch_size: [1,4,16]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1,2]
+            """)
+        # yapf: enable
+
+        yaml_config = self._load_config_file(yaml_str)
+        YamlConfigValidator.validate(yaml_config)
+
+    def test_invalid_yaml_file(self):
+        """
+        Raises an exception because run-config-search-max-instance-count: 16 uses
+        hyphens instead of the required underscores
+        """
+        # yapf: disable
+        yaml_str = ("""
+            run-config-search-max-instance-count: 16
+            run_config_search_disable: True
+            profile_models:
+                test_model:
+                    model_config_parameters:
+                        max_batch_size: [1,4,16]
+                        instance_group:
+                        -
+                            kind: KIND_GPU
+                            count: [1,2]
+            """)
+        # yapf: enable
+
+        yaml_config = self._load_config_file(yaml_str)
+        with self.assertRaises(TritonModelAnalyzerException):
+            YamlConfigValidator.validate(yaml_config)
+
+    def _load_config_file(self, yaml_str):
+        """
+        Load YAML config
+        """
+
+        config = yaml.safe_load(yaml_str)
+        return config

From fe1a251bbfddc44749d8733f8a5b007e19572ed1 Mon Sep 17 00:00:00 2001
From: David Yastremsky <dyastremsky@nvidia.com>
Date: Mon, 26 Feb 2024 15:45:50 -0800
Subject: [PATCH 2/4] Remove unused tests

---
 .../qa/L0_unit_tests/count_tests.py           |   57 +
 .../perf_analyzer/qa/L0_unit_tests/test.sh    |   50 +
 .../qa/common/check_analyzer_results.sh       |  115 +
 .../qa/common/check_copyright.py              |  239 ++
 src/c++/perf_analyzer/qa/common/util.sh       |  341 +++
 src/c++/perf_analyzer/tests/test_analyzer.py  |  241 --
 .../tests/test_bls_report_manager.py          |  172 --
 src/c++/perf_analyzer/tests/test_config.py    | 2382 -----------------
 .../tests/test_constraint_manager.py          |  620 -----
 .../perf_analyzer/tests/test_coordinate.py    |  138 -
 .../tests/test_coordinate_data.py             |  123 -
 .../perf_analyzer/tests/test_cpu_monitor.py   |  111 -
 .../perf_analyzer/tests/test_dcgm_monitor.py  |  155 --
 .../tests/test_ensemble_report_manager.py     |  178 --
 .../perf_analyzer/tests/test_file_writer.py   |   67 -
 .../perf_analyzer/tests/test_html_report.py   |  167 --
 src/c++/perf_analyzer/tests/test_logger.py    |  175 --
 .../perf_analyzer/tests/test_model_config.py  |  472 ----
 .../tests/test_model_config_generator.py      |  927 -------
 .../tests/test_model_config_measurement.py    |  242 --
 .../tests/test_model_constraints.py           |  119 -
 .../perf_analyzer/tests/test_model_manager.py | 1319 ---------
 .../tests/test_model_variant_name_manager.py  |  266 --
 .../perf_analyzer/tests/test_neighborhood.py  |  916 -------
 .../tests/test_parameter_search.py            |  336 ---
 .../perf_analyzer/tests/test_pdf_report.py    |   61 -
 src/c++/perf_analyzer/tests/test_plot.py      |  131 -
 .../perf_analyzer/tests/test_plot_manager.py  |  196 --
 .../tests/test_quick_run_config_generator.py  | 1359 ----------
 .../tests/test_record_aggregator.py           |  175 --
 .../perf_analyzer/tests/test_record_types.py  |  191 --
 .../tests/test_remote_monitor.py              |  174 --
 .../tests/test_report_manager.py              |  642 -----
 .../perf_analyzer/tests/test_report_utils.py  |   60 -
 .../tests/test_result_manager.py              |  199 --
 .../perf_analyzer/tests/test_result_table.py  |  121 -
 .../tests/test_result_table_manager.py        |  344 ---
 .../tests/test_run_config_generator.py        |  729 -----
 .../tests/test_run_config_measurement.py      |  739 -----
 .../test_run_config_result_comparator.py      |  258 --
 .../perf_analyzer/tests/test_search_config.py |   88 -
 .../tests/test_search_dimension.py            |   53 -
 .../tests/test_search_dimensions.py           |   79 -
 .../tests/test_sorted_results.py              |  306 ---
 .../perf_analyzer/tests/test_state_manager.py |  148 -
 .../perf_analyzer/tests/test_triton_client.py |  214 --
 .../perf_analyzer/tests/test_triton_server.py |  361 ---
 .../tests/test_triton_server_factory.py       |  126 -
 48 files changed, 802 insertions(+), 15880 deletions(-)
 create mode 100755 src/c++/perf_analyzer/qa/L0_unit_tests/count_tests.py
 create mode 100755 src/c++/perf_analyzer/qa/L0_unit_tests/test.sh
 create mode 100755 src/c++/perf_analyzer/qa/common/check_analyzer_results.sh
 create mode 100755 src/c++/perf_analyzer/qa/common/check_copyright.py
 create mode 100755 src/c++/perf_analyzer/qa/common/util.sh
 delete mode 100755 src/c++/perf_analyzer/tests/test_analyzer.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_bls_report_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_config.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_constraint_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_coordinate.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_coordinate_data.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_cpu_monitor.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_dcgm_monitor.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_ensemble_report_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_file_writer.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_html_report.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_logger.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_model_config.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_model_config_generator.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_model_config_measurement.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_model_constraints.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_model_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_model_variant_name_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_neighborhood.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_parameter_search.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_pdf_report.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_plot.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_plot_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_quick_run_config_generator.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_record_aggregator.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_record_types.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_remote_monitor.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_report_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_report_utils.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_result_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_result_table.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_result_table_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_run_config_generator.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_run_config_measurement.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_run_config_result_comparator.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_search_config.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_search_dimension.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_search_dimensions.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_sorted_results.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_state_manager.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_triton_client.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_triton_server.py
 delete mode 100755 src/c++/perf_analyzer/tests/test_triton_server_factory.py

diff --git a/src/c++/perf_analyzer/qa/L0_unit_tests/count_tests.py b/src/c++/perf_analyzer/qa/L0_unit_tests/count_tests.py
new file mode 100755
index 000000000..d8bf8315d
--- /dev/null
+++ b/src/c++/perf_analyzer/qa/L0_unit_tests/count_tests.py
@@ -0,0 +1,57 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import argparse
+import importlib
+import inspect
+import os
+import sys
+
+sys.path.insert(0, "../../")
+
+
+def args():
+    parser = argparse.ArgumentParser("test_counter")
+    parser.add_argument("--path", help="Path to use for counting the tests", type=str)
+    opt = parser.parse_args()
+    return opt
+
+
+if __name__ == "__main__":
+    number_of_tests = 0
+    opt = args()
+    path = opt.path
+
+    for file_path in os.listdir(path):
+        # All the test files start with "Test"
+        if file_path.startswith("test_"):
+            module_name = "tests." + file_path.split(".")[0]
+            module = importlib.import_module(module_name)
+            classes = inspect.getmembers(module, inspect.isclass)
+            for class_tuple in classes:
+                class_name = class_tuple[0]
+                class_object = class_tuple[1]
+
+                # All the test classes start with "Test"
+                if class_name.startswith("Test"):
+                    methods = inspect.getmembers(class_object, inspect.isroutine)
+                    for method_tuple in methods:
+                        method_name = method_tuple[0]
+                        if method_name.startswith("test_"):
+                            number_of_tests += 1
+
+    # Print the number of tests
+    print(number_of_tests)
diff --git a/src/c++/perf_analyzer/qa/L0_unit_tests/test.sh b/src/c++/perf_analyzer/qa/L0_unit_tests/test.sh
new file mode 100755
index 000000000..b664ea1d4
--- /dev/null
+++ b/src/c++/perf_analyzer/qa/L0_unit_tests/test.sh
@@ -0,0 +1,50 @@
+#!/bin/bash
+# Copyright (c) 2020-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+EXPECTED_NUM_TESTS=`python3 count_tests.py --path ../../tests/`
+source ../common/check_analyzer_results.sh
+source ../common/util.sh
+
+create_logs_dir "L0_unit_tests"
+create_result_paths -export-path false -checkpoints false
+
+RET=0
+
+set +e
+coverage run --branch --source=../../model_analyzer -m unittest discover -v -s ../../tests  -t ../../ > $ANALYZER_LOG 2>&1
+if [ $? -ne 0 ]; then
+    RET=1
+else
+    check_unit_test_results $ANALYZER_LOG $EXPECTED_NUM_TESTS
+    if [ $? -ne 0 ]; then
+        cat $ANALYZER_LOG
+        echo -e "\n***\n*** Test Result Verification Failed\n***"
+        RET=1
+    fi
+fi
+set -e
+
+if [ $RET -eq 0 ]; then
+    echo -e "\n***\n*** Test PASSED\n***"
+else
+    cat $ANALYZER_LOG
+    echo -e "\n***\n*** Test FAILED\n***"
+fi
+
+# Generate html files
+echo `pwd`
+coverage html --directory $LOGS_DIR/html/
+
+exit $RET
diff --git a/src/c++/perf_analyzer/qa/common/check_analyzer_results.sh b/src/c++/perf_analyzer/qa/common/check_analyzer_results.sh
new file mode 100755
index 000000000..79bc782d1
--- /dev/null
+++ b/src/c++/perf_analyzer/qa/common/check_analyzer_results.sh
@@ -0,0 +1,115 @@
+#!/bin/bash
+# Copyright (c) 2021-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+source ../common/util.sh
+
+# Check Python unittest results.
+function check_unit_test_results () {
+    local log_file=$1
+    local expected_num_tests=$2
+
+    if [[ -z "$expected_num_tests" ]]; then
+        echo "=== expected number of tests must be defined"
+        return 1
+    fi
+
+    local num_failures=`cat $log_file | grep -E ".*total.*errors.*failures.*" | tail -n 1 | jq .failures`
+    local num_tests=`cat $log_file | grep -E ".*total.*errors.*failures.*" | tail -n 1 | jq .total`
+    local num_errors=`cat $log_file | grep -E ".*total.*errors.*failures.*" | tail -n 1 | jq .errors`
+
+    # Number regular expression
+    re='^[0-9]+$'
+
+    if [[ $? -ne 0 ]] || ! [[ $num_failures =~ $re ]] || ! [[ $num_tests =~ $re ]] || \
+     ! [[ $num_errors =~ $re ]]; then
+        cat $log_file
+        echo -e "\n***\n*** Test Failed: unable to parse test results\n***" >> $log_file
+        return 1
+    fi
+    if [[ $num_errors != "0" ]] || [[ $num_failures != "0" ]] || [[ $num_tests -ne $expected_num_tests ]]; then
+        cat $log_file
+        echo -e "\n***\n*** Test Failed: Expected $expected_num_tests test(s), $num_tests test(s) executed, $num_errors test(s) had error, and $num_failures test(s) failed. \n***" >> $log_file
+        return 1
+    fi
+
+    return 0
+}
+
+# Check the table(s) by calling check_log_table_row_column and check_csv_table_row_column
+# See if the logic generalizes to the specify testing needs
+# Call check_log_table_row_column and check_csv_table_row_column in util.sh directly if not
+function check_table_row_column() {
+    local inference_log_file=$1
+    local gpu_log_file=$2
+    local server_log_file=$3
+    local inference_csv_file=$4
+    local gpu_csv_file=$5
+    local server_csv_file=$6
+
+    local inference_expected_num_columns=$7
+    local inference_expected_num_rows=$8
+    local gpu_expected_num_columns=$9
+    local gpu_expected_num_rows=${10}
+    local server_expected_num_columns=${11}
+    local server_expected_num_rows=${12}
+
+    local ret=0
+
+    if [ "$inference_log_file" != "" ]; then
+        check_log_table_row_column $inference_log_file $inference_expected_num_columns $inference_expected_num_rows "Models\ \(Inference\):"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $inference_log_file.\n***"
+            ret=1
+        fi
+    fi
+    if [ "$gpu_log_file" != "" ]; then
+        check_log_table_row_column $gpu_log_file $gpu_expected_num_columns $gpu_expected_num_rows "Models\ \(GPU\ Metrics\):"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $gpu_log_file.\n***"
+            ret=1
+        fi
+    fi
+    if [ "$server_log_file" != "" ]; then
+        check_log_table_row_column $server_log_file $server_expected_num_columns $server_expected_num_rows "Server\ Only:"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $server_log_file.\n***"
+            ret=1
+        fi
+    fi
+
+    if [ "$inference_csv_file" != "" ]; then
+        check_csv_table_row_column $inference_csv_file $inference_expected_num_columns $inference_expected_num_rows "Model"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $inference_csv_file.\n***"
+            ret=1
+        fi
+    fi
+    if [ "$gpu_csv_file" != "" ]; then
+        check_csv_table_row_column $gpu_csv_file $gpu_expected_num_columns $gpu_expected_num_rows "Model"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $gpu_csv_file.\n***"
+            ret=1
+        fi
+    fi
+    if [ "$server_csv_file" != "" ]; then
+        check_csv_table_row_column $server_csv_file $server_expected_num_columns $server_expected_num_rows "Model"
+        if [ $? -ne 0 ]; then
+            echo -e "\n***\n*** Test Output Verification Failed for $server_csv_file.\n***"
+            ret=1
+        fi
+    fi
+
+    return $ret
+}
diff --git a/src/c++/perf_analyzer/qa/common/check_copyright.py b/src/c++/perf_analyzer/qa/common/check_copyright.py
new file mode 100755
index 000000000..89c341512
--- /dev/null
+++ b/src/c++/perf_analyzer/qa/common/check_copyright.py
@@ -0,0 +1,239 @@
+#!/usr/bin/env python3
+
+# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import argparse
+import os
+import pathlib
+import re
+
+FLAGS = None
+SKIP_EXTS = ("pt", "log", "png", "pdf", "ckpt", "csv", "json")
+
+REPO_PATH_FROM_THIS_FILE = "../.."
+SKIP_PATHS = (".git", "VERSION", "LICENSE")
+
+COPYRIGHT_YEAR_RE = "Copyright( \\(c\\))? 20[1-9][0-9](-(20)?[1-9][0-9])?(,((20[2-9][0-9](-(20)?[2-9][0-9])?)|([2-9][0-9](-[2-9][0-9])?)))*,? NVIDIA CORPORATION( & AFFILIATES)?. All rights reserved."
+
+COPYRIGHT = """
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+"""
+
+repo_abs_path = (
+    pathlib.Path(__file__).parent.joinpath(REPO_PATH_FROM_THIS_FILE).resolve()
+)
+
+copyright_year_re = re.compile(COPYRIGHT_YEAR_RE)
+
+
+def visit(path):
+    if FLAGS.verbose:
+        print("visiting " + path)
+
+    for skip in SKIP_EXTS:
+        if path.endswith("." + skip):
+            if FLAGS.verbose:
+                print("skipping due to extension: " + path)
+            return True
+
+    for skip in SKIP_PATHS:
+        if str(pathlib.Path(path).resolve()).startswith(
+            str(repo_abs_path.joinpath(skip).resolve())
+        ):
+            if FLAGS.verbose:
+                print("skipping due to path prefix: " + path)
+            return True
+
+    with open(path, "r") as f:
+        first_line = True
+        line = None
+        try:
+            for fline in f:
+                line = fline
+
+                # Skip any '#!', '..', '<!--', or '{{/*' lines at the
+                # start of the file
+                if first_line:
+                    first_line = False
+                    if (
+                        fline.startswith("#!")
+                        or fline.startswith("..")
+                        or fline.startswith("<!--")
+                        or fline.startswith("{{/*")
+                    ):
+                        continue
+                # Skip empty lines...
+                if len(fline.strip()) != 0:
+                    break
+        except UnicodeDecodeError as ex:
+            # If we get this exception on the first line then assume a
+            # non-text file.
+            if not first_line:
+                raise ex
+            if FLAGS.verbose:
+                print("skipping binary file: " + path)
+            return True
+        if line is None:
+            if FLAGS.verbose:
+                print("skipping empty file: " + path)
+            return True
+
+        line = line.strip()
+
+        # The next line must be the copyright line with a single year
+        # or a year range. It is optionally allowed to have '# ' or
+        # '// ' prefix.
+        prefix = ""
+        if line.startswith("# "):
+            prefix = "# "
+        elif line.startswith("// "):
+            prefix = "// "
+        elif not line.startswith(COPYRIGHT_YEAR_RE[0]):
+            print(
+                "incorrect prefix for copyright line, allowed prefixes '# ' or '// ', for "
+                + path
+                + ": "
+                + line
+            )
+            return False
+
+        # Check if the copyright year line matches the regex
+        # and see if the year(s) are reasonable
+        years = []
+
+        copyright_row = line[len(prefix) :]
+        if copyright_year_re.match(copyright_row):
+            for year in (
+                copyright_row.split(
+                    "(c) " if "(c) " in copyright_row else "Copyright "
+                )[1]
+                .split(" NVIDIA ")[0]
+                .split(",")
+            ):
+                if len(year) == 4:  # 2021
+                    years.append(int(year))
+                elif len(year) == 2:  # 21
+                    years.append(int(year) + 2000)
+                elif len(year) == 9:  # 2021-2022
+                    years.append(int(year[0:4]))
+                    years.append(int(year[5:9]))
+                elif len(year) == 7:  # 2021-22
+                    years.append(int(year[0:4]))
+                    years.append(int(year[5:7]) + 2000)
+                elif len(year) == 5:  # 21-23
+                    years.append(int(year[0:2]) + 2000)
+                    years.append(int(year[3:5]) + 2000)
+        else:
+            print("copyright year is not recognized for " + path + ": " + line)
+            return False
+
+        if years[0] > FLAGS.year:
+            print(
+                "copyright start year greater than current year for "
+                + path
+                + ": "
+                + line
+            )
+            return False
+        if years[-1] > FLAGS.year:
+            print(
+                "copyright end year greater than current year for " + path + ": " + line
+            )
+            return False
+        for i in range(1, len(years)):
+            if years[i - 1] >= years[i]:
+                print("copyright years are not increasing for " + path + ": " + line)
+                return False
+
+        # Subsequent lines must match the copyright body.
+        copyright_body = [
+            l.rstrip() for i, l in enumerate(COPYRIGHT.splitlines()) if i > 0
+        ]
+        copyright_idx = 0
+        for line in f:
+            if copyright_idx >= len(copyright_body):
+                break
+
+            if len(prefix) == 0:
+                line = line.rstrip()
+            else:
+                line = line.strip()
+
+            if len(copyright_body[copyright_idx]) == 0:
+                expected = prefix.strip()
+            else:
+                expected = prefix + copyright_body[copyright_idx]
+            if line != expected:
+                print("incorrect copyright body for " + path)
+                print("  expected: '" + expected + "'")
+                print("       got: '" + line + "'")
+                return False
+            copyright_idx += 1
+
+        if copyright_idx != len(copyright_body):
+            print(
+                "missing "
+                + str(len(copyright_body) - copyright_idx)
+                + " lines of the copyright body"
+            )
+            return False
+
+    if FLAGS.verbose:
+        print("copyright correct for " + path)
+    return True
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "-v",
+        "--verbose",
+        action="store_true",
+        required=False,
+        default=False,
+        help="Enable verbose output",
+    )
+    parser.add_argument("-y", "--year", type=int, required=True, help="Copyright year")
+    parser.add_argument(
+        "paths", type=str, nargs="*", default=None, help="Directories or files to check"
+    )
+    FLAGS = parser.parse_args()
+
+    if FLAGS.paths is None or len(FLAGS.paths) == 0:
+        parser.print_help()
+        exit(1)
+
+    ret = True
+    for path in FLAGS.paths:
+        if not os.path.isdir(path):
+            if not visit(path):
+                ret = False
+        else:
+            for root, dirs, files in os.walk(path):
+                for name in files:
+                    if not visit(os.path.join(root, name)):
+                        ret = False
+
+    exit(0 if ret else 1)
diff --git a/src/c++/perf_analyzer/qa/common/util.sh b/src/c++/perf_analyzer/qa/common/util.sh
new file mode 100755
index 000000000..84054858a
--- /dev/null
+++ b/src/c++/perf_analyzer/qa/common/util.sh
@@ -0,0 +1,341 @@
+#!/bin/bash
+# Copyright (c) 2020-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+LOGS_DIR=${LOGS_DIR:="/logs"}
+SERVER_LOG=${SERVER_LOG:="$LOGS_DIR/server.log"}
+SERVER_TIMEOUT=${SERVER_TIMEOUT:=120}
+SERVER_HTTP_PORT=${SERVER_HTTP_PORT:=8000}
+SERVER_LD_PRELOAD=${SERVER_LD_PRELOAD:=""}
+ANALYZER_LOG=${ANALYZER_LOG:="$LOGS_DIR/test.log"}
+
+mkdir -p $LOGS_DIR
+
+function create_logs_dir() {
+    # Arguments:
+    #  $1: L0 Script name
+    # Check if the L0 script name is empty or not
+    if [ -n "$1" ]; then
+        LOGS_DIR="/logs/$1"
+    else
+        LOGS_DIR="/logs"
+    fi
+    mkdir -p "$LOGS_DIR"
+}
+
+function create_result_paths() {
+    # Creates ANALYZER_LOG, EXPORT_PATH, CHECKPOINT_DIRECTORY and TEST_LOG_DIR
+    # Arguments:
+    #  -test-name <test_name>: <string> - L0 Script name (optional argument)
+    #  -export-path <false>: <boolean> - If false, skip creating EXPORT_PATH (optional argument)
+    #  -checkpoints <false>: <boolean> - If false, skip creating CHECKPOINT_DIRECTORY (optional argument)
+
+    # Set default values
+    test_name=""
+    export_path=true
+    checkpoints=true
+
+    # Parse arguments options
+    while [[ $# -gt 0 ]]; do
+        case $1 in
+            -test-name)
+                test_name=$2;
+                shift 2;;
+            -export-path)
+                export_path=$2;
+                shift 2;;
+            -checkpoints)
+                checkpoints=$2;
+                shift 2;;
+            *)
+                echo "Invalid option: $1" >&2;
+                return 1;;
+        esac
+    done
+
+    # Check if the test name is not an empty string
+    if [ -n "$test_name" ]; then
+        TEST_LOG_DIR="$LOGS_DIR/$test_name/logs"
+        ANALYZER_LOG="$TEST_LOG_DIR/analyzer.${test_name}.log"
+        mkdir -p "$TEST_LOG_DIR"
+
+        # Create EXPORT_PATH if export_path is true
+        if [ "$export_path" = true ]; then
+            EXPORT_PATH="$LOGS_DIR/$test_name/results"
+            mkdir -p "$EXPORT_PATH"
+        fi
+
+        # Create CHECKPOINT_DIRECTORY if checkpoints is true
+        if [ "$checkpoints" = true ]; then
+            CHECKPOINT_DIRECTORY="$LOGS_DIR/$test_name/checkpoints"
+            mkdir -p "$CHECKPOINT_DIRECTORY"
+        fi
+    else
+        TEST_LOG_DIR="$LOGS_DIR/logs"
+        ANALYZER_LOG="$TEST_LOG_DIR/analyzer.log"
+        mkdir -p "$TEST_LOG_DIR"
+
+        # Create EXPORT_PATH if export_path is true
+        if [ "$export_path" = true ]; then
+            EXPORT_PATH="$LOGS_DIR/results"
+            mkdir -p "$EXPORT_PATH"
+        fi
+
+        # Create CHECKPOINT_DIRECTORY if checkpoints is true
+        if [ "$checkpoints" = true ]; then
+            CHECKPOINT_DIRECTORY="$LOGS_DIR/checkpoints"
+            mkdir -p "$CHECKPOINT_DIRECTORY"
+        fi
+    fi
+}
+
+# Wait until server health endpoint shows ready. Sets WAIT_RET to 0 on
+# success, 1 on failure
+function wait_for_server_ready() {
+    local spid="$1"; shift
+    local wait_time_secs="${1:-30}"; shift
+
+    WAIT_RET=0
+
+    local wait_secs=$wait_time_secs
+    until test $wait_secs -eq 0 ; do
+        if ! kill -0 $spid; then
+            echo "=== Server not running."
+            WAIT_RET=1
+            return
+        fi
+
+        sleep 1;
+
+        set +e
+        code=`curl -s -w %{http_code} localhost:${1}/v2/health/ready`
+        set -e
+        if [ "$code" == "200" ]; then
+            return
+        fi
+
+        ((wait_secs--));
+    done
+
+    echo "=== Timeout $wait_time_secs secs. Server not ready."
+    WAIT_RET=1
+}
+
+# Run inference server. Return once server's health endpoint shows
+# ready or timeout expires. Sets SERVER_PID to pid of SERVER, or 0 if
+# error (including expired timeout)
+function run_server () {
+    SERVER_PID=0
+
+    if [ -z "$SERVER" ]; then
+        echo "=== SERVER must be defined"
+        return
+    fi
+
+    if [ ! -f "$SERVER" ]; then
+        echo "=== $SERVER does not exist"
+        return
+    fi
+
+    if [ -z "$SERVER_LD_PRELOAD" ]; then
+      echo "=== Running $SERVER $SERVER_ARGS"
+    else
+      echo "=== Running LD_PRELOAD=$SERVER_LD_PRELOAD $SERVER $SERVER_ARGS"
+    fi
+
+    LD_PRELOAD=$SERVER_LD_PRELOAD $SERVER $SERVER_ARGS > $SERVER_LOG 2>&1 &
+    SERVER_PID=$!
+
+    wait_for_server_ready $SERVER_PID $SERVER_TIMEOUT $SERVER_HTTP_PORT
+    if [ "$WAIT_RET" != "0" ]; then
+        kill $SERVER_PID || true
+        SERVER_PID=0
+    fi
+}
+
+# Run model-analyzer with args to completion.
+function run_analyzer() {
+    if [[ -z "$MODEL_ANALYZER" ]]; then
+        echo -e "=== model-analyzer executable not found"
+        return 1
+    fi
+
+    if [ ! -f "$MODEL_ANALYZER" ]; then
+        echo "=== $MODEL_ANALYZER does not exist"
+        return 1
+    fi
+
+    echo -e "=== Running $MODEL_ANALYZER $MODEL_ANALYZER_GLOBAL_OPTIONS $MODEL_ANALYZER_SUBCOMMAND $MODEL_ANALYZER_ARGS"
+    $MODEL_ANALYZER $MODEL_ANALYZER_GLOBAL_OPTIONS $MODEL_ANALYZER_SUBCOMMAND $MODEL_ANALYZER_ARGS >> $ANALYZER_LOG 2>&1
+    return $?
+}
+
+# Run model-analyzer with args to completion.
+function run_analyzer_nohup() {
+    if [[ -z "$MODEL_ANALYZER" ]]; then
+        echo -e "=== model-analyzer executable not found"
+        return 1
+    fi
+
+    if [ ! -f "$MODEL_ANALYZER" ]; then
+        echo "=== $MODEL_ANALYZER does not exist"
+        return 1
+    fi
+
+    echo -e "=== Running $MODEL_ANALYZER $MODEL_ANALYZER_GLOBAL_OPTIONS $MODEL_ANALYZER_SUBCOMMAND $MODEL_ANALYZER_ARGS"
+    $MODEL_ANALYZER $MODEL_ANALYZER_GLOBAL_OPTIONS $MODEL_ANALYZER_SUBCOMMAND $MODEL_ANALYZER_ARGS >> $ANALYZER_LOG 2>&1 &
+}
+
+# Check row and columns of csv file
+function check_csv_table_row_column() {
+    local csv_file=$1
+    local expected_num_columns=$2
+    local expected_num_rows=$3
+    local tag=$4
+    if [[ ! -f "$csv_file" ]]; then
+        echo -e "\n***\n*** Test Failed: $csv_file does not exist\n***"
+        return 1
+    fi
+
+    num_rows=`awk -v pattern=$tag '$0 ~ pattern {getline; i=0; while(getline) {i+=1}; print i}' $csv_file`
+    if [[ "$num_rows" != "$expected_num_rows" ]]; then
+        echo -e "\n***\n*** Test Failed: Expected $expected_num_rows rows in $csv_file, got ${num_rows}\n***"
+        echo -e "csv tag = ${tag}"
+        echo -e "csv csv_file = ${csv_file}"
+        echo -e "vvvvvvvvv"
+        cat ${csv_file}
+        echo -e "^^^^^^^^^"
+        return 1
+    fi
+    for i in $( seq 1 $expected_num_rows ); do
+        num_columns_found=`awk -v pattern=$tag -F ',' -v row="$i" '$0 ~ pattern {for (n=0; n<row;n++) {getline}; print NF}' $csv_file`
+        if [[ "$num_columns_found" != "$expected_num_columns" ]]; then
+            echo -e "\n***\n*** Test Failed: Expected $expected_num_columns columns in row $i, got ${num_columns_found}\n***"
+            echo -e "csv tag = ${tag}"
+            echo -e "csv row = ${i}"
+            echo -e "csv csv_file = ${csv_file}"
+            echo -e "vvvvvvvvv"
+            cat ${csv_file}
+            echo -e "^^^^^^^^^"
+            return 1
+        fi
+    done
+    return 0
+}
+
+function check_no_csv_exists() {
+    local csv_file=$1
+    if [[ ! -f "$csv_file" ]]; then
+        return 0
+    else
+        echo -e "\n***\n*** Test Failed: $csv_file should not exist\n***"
+        return 1
+    fi
+}
+
+function get_number_of_rows_logfile() {
+    local log_file=$1
+    num_rows_found=`awk  "BEGIN{i=0} /$tag/{flag=1;getline;getline} /^$/{flag=0} flag {i+=1} END{print i}" $log_file`
+    echo $number_row_found
+}
+
+function get_number_of_rows_csv() {
+    local csv_file=$1
+    num_rows=`awk -v pattern=$tag '$0 ~ pattern {getline; i=0; while(getline) {i+=1}; print i}' $csv_file`
+    echo $num_rows
+}
+
+# Check the output tables from the model-analyzer
+# This function simply ensures that there is a
+# Server-only table, a model table, and that
+# They have the specified dimensions
+function check_log_table_row_column() {
+    local log_file=$1
+    local expected_num_columns=$2
+    local expected_num_rows=$3
+    local tag=$4
+
+    # Check number of rows
+    num_rows_found=`awk  "BEGIN{i=0} /$tag/{flag=1;getline;getline} /^$/{flag=0} flag {i+=1} END{print i}" $log_file`
+    if [[ "$num_rows_found" != "$expected_num_rows" ]]; then
+        echo -e "\n***\n*** Test Failed: Expected $expected_num_rows rows $log_file, got ${num_rows_found}\n***"
+        echo -e "log tag = ${tag}"
+        echo -e "log log_file = ${log_file}"
+        echo -e "vvvvvvvvv"
+        cat ${log_file}
+        echo -e "^^^^^^^^^"
+        return 1
+    fi
+
+    # Check models table
+    for i in $( seq 1 $expected_num_rows ); do
+        # Columns in ith row
+        num_columns_found=`awk -v row="$i" "/$tag/{getline; for (n=0; n<row;n++) {getline}; print NF}" $log_file`
+        if [[ "$num_columns_found" != "$expected_num_columns" ]]; then
+            echo -e "\n***\n*** Test Failed: Expected $expected_num_columns columns in row $i of $tag, got ${num_columns_found}\n***"
+            echo -e "log tag = ${tag}"
+            echo -e "log row = ${i}"
+            echo -e "log log_file = ${log_file}"
+            echo -e "vvvvvvvvv"
+            cat ${log_file}
+            echo -e "^^^^^^^^^"
+            return 1
+        fi
+    done
+    return 0
+}
+
+function install_netstat() {
+    netstat > /dev/null 2>&1
+    if [ $? -eq 0 ]; then
+        return
+    else
+        DEBIAN_FRONTEND=noninteractive
+        apt-get update -qq > /dev/null 2>&1
+        DEBIAN_FRONTEND=noninteractive
+        apt-get install net-tools -y -qq > /dev/null 2>&1
+    fi
+}
+
+function find_available_ports() {
+    install_netstat
+    # First argument is the number of ports
+    num_port=$1
+
+    export ports=()
+    for i in `seq 1 $num_port`; do
+        current_port=$((10000 + $RANDOM % 10000))
+        while [ `netstat -ano tcp |& grep :$current_port > /dev/null 2>&1` ]; do
+            current_port=$(echo "$port + 1" | bc)
+        done
+        ports=("${ports[@]}" "$current_port")
+    done
+    echo ${ports[@]}
+}
+
+# Get a space delimited list of gpus with their UUIDs.
+function get_all_gpus_uuids() {
+    nvidia-smi --query-gpu=gpu_uuid,pci.bus_id --format=csv,noheader | awk 'BEGIN { FS = "," }{print $1}'
+}
+
+# We need to randomize the output directory so that in case the CI containers
+# end up on the same host, it does not create a conflict
+function get_output_directory() {
+    random_number=$((10000 + $RANDOM % 10000))
+    while [ `ls /tmp/output/ | grep $random_number` ]; do
+        random_number=$(echo "$random_number + 1" | bc)
+    done
+
+    echo /tmp/output/$random_number
+}
diff --git a/src/c++/perf_analyzer/tests/test_analyzer.py b/src/c++/perf_analyzer/tests/test_analyzer.py
deleted file mode 100755
index 68680ff57..000000000
--- a/src/c++/perf_analyzer/tests/test_analyzer.py
+++ /dev/null
@@ -1,241 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.analyzer import Analyzer
-from model_analyzer.config.input.config_status import ConfigStatus
-from model_analyzer.config.run.model_run_config import ModelRunConfig
-from model_analyzer.config.run.run_config import RunConfig
-from model_analyzer.constants import CONFIG_PARSER_SUCCESS
-from model_analyzer.result.results import Results
-from model_analyzer.result.run_config_result import RunConfigResult
-from model_analyzer.result.sorted_results import SortedResults
-from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
-from model_analyzer.triton.model.model_config import ModelConfig
-from model_analyzer.triton.model.model_config_variant import ModelConfigVariant
-from tests.common.test_utils import evaluate_mock_config
-
-from .common import test_result_collector as trc
-
-
-class TestAnalyzer(trc.TestResultCollector):
-    """
-    Tests the methods of the Analyzer class
-    """
-
-    def tearDown(self):
-        patch.stopall()
-
-    def mock_get_state_variable(self, name):
-        if name == "ResultManager.results":
-            return Results()
-        else:
-            return {
-                "model1": {
-                    "config1": None,
-                    "config2": None,
-                    "config3": None,
-                    "config4": None,
-                }
-            }
-
-    def mock_get_list_of_models(self):
-        return ["model1"]
-
-    @patch.multiple(
-        f"{AnalyzerStateManager.__module__}.AnalyzerStateManager",
-        get_state_variable=mock_get_state_variable,
-        exiting=lambda _: False,
-    )
-    @patch.multiple(
-        f"{Analyzer.__module__}.Analyzer",
-        _create_metrics_manager=MagicMock(),
-        _create_model_manager=MagicMock(),
-        _get_server_only_metrics=MagicMock(),
-        _profile_models=MagicMock(),
-        _check_for_perf_analyzer_errors=MagicMock(),
-    )
-    def test_profile_skip_summary_reports(self, **mocks):
-        """
-        Tests when the skip_summary_reports config option is turned on,
-        the profile stage does not create any summary reports.
-
-        NOTE: this test only ensures that the reports are not created with
-        the default export-path.
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "/tmp",
-            "--profile-models",
-            "model1",
-            "--config-file",
-            "/tmp/my_config.yml",
-            "--checkpoint-directory",
-            "/tmp/my_checkpoints",
-            "--skip-summary-reports",
-        ]
-        config = evaluate_mock_config(args, "", subcommand="profile")
-        state_manager = AnalyzerStateManager(config, None)
-        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
-        analyzer.profile(client=None, gpus=None, mode=None, verbose=False)
-
-        path = os.getcwd()
-        self.assertFalse(os.path.exists(os.path.join(path, "plots")))
-        self.assertFalse(os.path.exists(os.path.join(path, "results")))
-        self.assertFalse(os.path.exists(os.path.join(path, "reports")))
-
-    def mock_top_n_results(self, model_name=None, n=SortedResults.GET_ALL_RESULTS):
-        rc1 = RunConfig({})
-        rc1.add_model_run_config(
-            ModelRunConfig(
-                "fake_model_name",
-                ModelConfigVariant(
-                    ModelConfig.create_from_dictionary({"name": "config1"}),
-                    "config1",
-                ),
-                MagicMock(),
-            )
-        )
-        rc2 = RunConfig({})
-        rc2.add_model_run_config(
-            ModelRunConfig(
-                "fake_model_name",
-                ModelConfigVariant(
-                    ModelConfig.create_from_dictionary({"name": "config3"}),
-                    "config3",
-                ),
-                MagicMock(),
-            )
-        )
-        rc3 = RunConfig({})
-        rc3.add_model_run_config(
-            ModelRunConfig(
-                "fake_model_name",
-                ModelConfigVariant(
-                    ModelConfig.create_from_dictionary({"name": "config4"}),
-                    "config4",
-                ),
-                MagicMock(),
-            )
-        )
-
-        return [
-            RunConfigResult("fake_model_name", rc1, MagicMock(), MagicMock()),
-            RunConfigResult("fake_model_name", rc2, MagicMock(), MagicMock()),
-            RunConfigResult("fake_model_name", rc3, MagicMock(), MagicMock()),
-        ]
-
-    @patch(
-        "model_analyzer.config.input.config_command_profile.file_path_validator",
-        lambda _: ConfigStatus(status=CONFIG_PARSER_SUCCESS),
-    )
-    @patch(
-        "model_analyzer.config.input.config_command_profile.ConfigCommandProfile._preprocess_and_verify_arguments",
-        lambda _: None,
-    )
-    @patch(
-        "model_analyzer.result.result_manager.ResultManager.top_n_results",
-        mock_top_n_results,
-    )
-    def test_get_report_command_help_string(self):
-        """
-        Tests that the member function returning the report command help string
-        works correctly.
-        """
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "--profile-models",
-            "model1",
-            "--config-file",
-            "/tmp/my_config.yml",
-            "--checkpoint-directory",
-            "/tmp/my_checkpoints",
-            "--export-path",
-            "/tmp/my_export_path",
-            "--model-repository",
-            "cli_repository",
-        ]
-        config = evaluate_mock_config(args, "", subcommand="profile")
-        state_manager = AnalyzerStateManager(config, None)
-        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
-        self.assertEqual(
-            analyzer._get_report_command_help_string(model_name="model1"),
-            "To generate detailed reports for the 3 best model1 configurations, run "
-            "`model-analyzer report --report-model-configs "
-            "config1,config3,config4 --export-path /tmp/my_export_path "
-            "--config-file /tmp/my_config.yml --checkpoint-directory "
-            "/tmp/my_checkpoints`",
-        )
-
-    def test_multiple_models_in_report_model_config(self):
-        """
-        Test that multiple different models in the report are detected
-        """
-        # Single model config
-        args = [
-            "model-analyzer",
-            "report",
-            "--report-model-configs",
-            "modelA_config_0",
-            "--checkpoint-directory",
-            "/tmp/my_checkpoints",
-        ]
-        config = evaluate_mock_config(args, "", subcommand="report")
-        state_manager = AnalyzerStateManager(config, None)
-        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
-
-        self.assertFalse(analyzer._multiple_models_in_report_model_config())
-
-        # Multiple different models
-        args = [
-            "model-analyzer",
-            "report",
-            "--report-model-configs",
-            "modelA_config_0,modelB_config_1",
-            "--checkpoint-directory",
-            "/tmp/my_checkpoints",
-        ]
-        config = evaluate_mock_config(args, "", subcommand="report")
-        state_manager = AnalyzerStateManager(config, None)
-        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
-
-        self.assertTrue(analyzer._multiple_models_in_report_model_config())
-
-        # Single model - multiple configs
-        args = [
-            "model-analyzer",
-            "report",
-            "--report-model-configs",
-            "modelA_config_0,modelA_config_1",
-            "--checkpoint-directory",
-            "/tmp/my_checkpoints",
-        ]
-        config = evaluate_mock_config(args, "", subcommand="report")
-        state_manager = AnalyzerStateManager(config, None)
-        analyzer = Analyzer(config, None, state_manager, checkpoint_required=False)
-
-        self.assertFalse(analyzer._multiple_models_in_report_model_config())
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_bls_report_manager.py b/src/c++/perf_analyzer/tests/test_bls_report_manager.py
deleted file mode 100755
index 911369081..000000000
--- a/src/c++/perf_analyzer/tests/test_bls_report_manager.py
+++ /dev/null
@@ -1,172 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import shutil
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.reports.report_manager import ReportManager
-from model_analyzer.result.result_manager import ResultManager
-from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
-
-from .common import test_result_collector as trc
-from .common.test_utils import ROOT_DIR, evaluate_mock_config
-
-
-class TestBLSReportManager(trc.TestResultCollector):
-    def _init_managers(
-        self,
-        models="test_model",
-        num_configs_per_model=10,
-        mode="online",
-        subcommand="profile",
-    ):
-        args = ["model-analyzer", subcommand, "-f", "path-to-config-file"]
-        args.extend(["--checkpoint-directory", f"{ROOT_DIR}/bls-ckpt"])
-
-        if subcommand == "profile":
-            args.extend(["--profile-models", models])
-            args.extend(["--model-repository", "/tmp"])
-        else:
-            args.extend(["--report-model-configs", models])
-
-        yaml_str = f"""
-            client_protocol: grpc
-            export_path: {ROOT_DIR}
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand=subcommand)
-        state_manager = AnalyzerStateManager(config=config, server=None)
-
-        state_manager.load_checkpoint(checkpoint_required=True)
-        gpu_info = {"gpu_uuid": {"name": "fake_gpu_name", "total_memory": 1024000000}}
-        self.result_manager = ResultManager(
-            config=config, state_manager=state_manager, constraint_manager=MagicMock()
-        )
-        self.report_manager = ReportManager(
-            mode=mode,
-            config=config,
-            gpu_info=gpu_info,
-            result_manager=self.result_manager,
-            constraint_manager=MagicMock(),
-        )
-
-    def test_bls_summary(self):
-        """
-        Ensures the summary report sentence and table are accurate for a basic bls model (loaded from a checkpoint)
-        """
-        self._init_managers(models="bls", subcommand="profile")
-
-        self.report_manager.create_summaries()
-
-        expected_summary_sentence = (
-            "In 54 measurements across 9 configurations, <strong>bls_config_6</strong> "
-            "is <strong>1111%</strong> better than the default configuration at maximizing throughput, "
-            "under the given constraints, on GPU(s) TITAN RTX.<UL><LI> <strong>bls_config_6</strong>: 8 GPU instances with a max batch size of 0 on platform python "
-            "</LI><BR>Which is comprised of the following composing models: "
-            "<UL><LI> <strong>add_config_4</strong>: 6 GPU instances with a max batch size of 0 on platform python "
-            "</LI><LI> <strong>sub_config_0</strong>: 1 GPU instance with a max batch size of 0 on platform python </LI> </UL>"
-        )
-
-        summary_table, summary_sentence = self.report_manager._build_summary_table(
-            report_key="bls",
-            num_measurements=54,
-            num_configurations=9,
-            gpu_name="TITAN RTX",
-            report_gpu_metrics=True,
-        )
-
-        self.assertEqual(summary_sentence, expected_summary_sentence)
-
-        # Check the first row of the table
-        self.assertEqual(summary_table._rows[0][0], "bls_config_6")  # model config name
-        self.assertEqual(summary_table._rows[0][1], "(0, 0)")  # max batch size
-        self.assertEqual(
-            summary_table._rows[0][2], "(Disabled, Disabled)"
-        )  # dynamic batching
-        self.assertEqual(summary_table._rows[0][3], "(6:GPU, 1:GPU)")  # instance count
-        self.assertEqual(summary_table._rows[0][4], "2.215")  # p99 latency
-        self.assertEqual(summary_table._rows[0][5], "10077.3")  # throughput
-        self.assertEqual(summary_table._rows[0][6], 874)  # max gpu memory
-        self.assertEqual(summary_table._rows[0][7], 0.0)  # GPU utilization
-
-    def test_bls_summary_cpu_only(self):
-        """
-        Ensures the summary report sentence and table are accurate for a basic BLS model (loaded from a checkpoint)
-        when the cpu only flag is set
-        """
-
-        self._init_managers(models="bls", subcommand="profile")
-
-        self.report_manager.create_summaries()
-
-        expected_summary_sentence = (
-            "In 54 measurements across 9 configurations, <strong>bls_config_6</strong> "
-            "is <strong>1111%</strong> better than the default configuration at maximizing throughput, "
-            "under the given constraints.<UL><LI> <strong>bls_config_6</strong>: 8 GPU instances with a max batch size of 0 on platform python "
-            "</LI><BR>Which is comprised of the following composing models: "
-            "<UL><LI> <strong>add_config_4</strong>: 6 GPU instances with a max batch size of 0 on platform python "
-            "</LI><LI> <strong>sub_config_0</strong>: 1 GPU instance with a max batch size of 0 on platform python </LI> </UL>"
-        )
-
-        summary_table, summary_sentence = self.report_manager._build_summary_table(
-            report_key="bls",
-            num_measurements=54,
-            num_configurations=9,
-            gpu_name="TITAN RTX",
-            report_gpu_metrics=False,
-        )
-
-        self.assertEqual(summary_sentence, expected_summary_sentence)
-
-        # Check the first row of the table
-        self.assertEqual(summary_table._rows[0][0], "bls_config_6")  # model config name
-        self.assertEqual(summary_table._rows[0][1], "0, 0")  # max batch size
-        self.assertEqual(
-            summary_table._rows[0][2], "(Disabled, Disabled)"
-        )  # dynamic batching
-        self.assertEqual(summary_table._rows[0][3], "6:GPU, 1:GPU")  # instance count
-        self.assertEqual(summary_table._rows[0][4], "2.215")  # p99 latency
-        self.assertEqual(summary_table._rows[0][5], "10077.3")  # throughput
-
-    def test_bls_detailed(self):
-        """
-        Ensures the detailed report sentence is accurate for a BLS model (loaded from a checkpoint)
-        """
-        self._init_managers(models="bls_config_6", subcommand="report")
-
-        self.report_manager._add_detailed_report_data()
-        self.report_manager._build_detailed_table("bls_config_6")
-        detailed_sentence = self.report_manager._build_detailed_info("bls_config_6")
-
-        expected_detailed_sentence = (
-            "<strong>bls_config_6</strong> is comprised of the "
-            "following composing models:<LI> <strong>add_config_4</strong>: "
-            "6 GPU instances with a max batch size of 0 on platform python </LI><LI> "
-            "<strong>sub_config_0</strong>: 1 GPU instance with a max batch size of 0 "
-            "on platform python </LI><br>12 measurement(s) "
-            "were obtained for the model config on GPU(s)  with total memory 0 GB."
-        )
-
-        self.assertEqual(detailed_sentence, expected_detailed_sentence)
-
-    def tearDown(self):
-        patch.stopall()
-        shutil.rmtree(f"{ROOT_DIR}/reports")
-        shutil.rmtree(f"{ROOT_DIR}/plots")
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_config.py b/src/c++/perf_analyzer/tests/test_config.py
deleted file mode 100755
index f056eb76a..000000000
--- a/src/c++/perf_analyzer/tests/test_config.py
+++ /dev/null
@@ -1,2382 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import re
-import unittest
-from argparse import Namespace
-from copy import deepcopy
-from typing import Any, Dict, List, Optional
-from unittest.mock import patch
-
-from model_analyzer.cli.cli import CLI
-from model_analyzer.config.input.config_command_profile import ConfigCommandProfile
-from model_analyzer.config.input.config_command_report import ConfigCommandReport
-from model_analyzer.config.input.config_enum import ConfigEnum
-from model_analyzer.config.input.config_list_generic import ConfigListGeneric
-from model_analyzer.config.input.config_list_numeric import ConfigListNumeric
-from model_analyzer.config.input.config_list_string import ConfigListString
-from model_analyzer.config.input.config_object import ConfigObject
-from model_analyzer.config.input.config_primitive import ConfigPrimitive
-from model_analyzer.config.input.config_sweep import ConfigSweep
-from model_analyzer.config.input.config_union import ConfigUnion
-from model_analyzer.config.input.objects.config_model_profile_spec import (
-    ConfigModelProfileSpec,
-)
-from model_analyzer.config.input.objects.config_plot import ConfigPlot
-from model_analyzer.constants import CONFIG_PARSER_FAILURE
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.result.model_constraints import ModelConstraints
-
-from .common import test_result_collector as trc
-from .mocks.mock_config import MockConfig
-from .mocks.mock_numba import MockNumba
-from .mocks.mock_os import MockOSMethods
-
-
-class TestConfig(trc.TestResultCollector):
-    def _create_parameters(
-        self, batch_sizes: List = [], concurrency: List = [], request_rate: List = []
-    ) -> Dict:
-        return {
-            "batch_sizes": batch_sizes,
-            "concurrency": concurrency,
-            "request_rate": request_rate,
-        }
-
-    def _evaluate_config(
-        self, args, yaml_content, subcommand="profile", numba_available=True
-    ):
-        mock_numba = MockNumba(
-            mock_paths=["model_analyzer.config.input.config_command_profile"],
-            is_available=numba_available,
-        )
-
-        mock_config = MockConfig(args, yaml_content)
-        mock_config.start()
-        mock_numba.start()
-
-        if subcommand == "report":
-            config = ConfigCommandReport()
-        else:
-            config = ConfigCommandProfile()
-        cli = CLI()
-        cli.add_subcommand(cmd=subcommand, config=config, help="Test subcommand help")
-        cli.parse()
-        mock_config.stop()
-        mock_numba.stop()
-        return config
-
-    def _assert_error_on_evaluate_config(
-        self, args, yaml_content, subcommand="profile"
-    ):
-        mock_config = MockConfig(args, yaml_content)
-        mock_config.start()
-        if subcommand == "report":
-            config = ConfigCommandReport()
-        else:
-            config = ConfigCommandProfile()
-        cli = CLI()
-        cli.add_subcommand(cmd=subcommand, config=config, help="Test subcommand help")
-        # When a required field is not specified, parse will lead to an
-        # exception
-        with self.assertRaises(TritonModelAnalyzerException):
-            cli.parse()
-        mock_config.stop()
-
-    def _assert_equality_of_model_configs(self, model_configs, expected_model_configs):
-        self.assertEqual(len(model_configs), len(expected_model_configs))
-        for model_config, expected_model_config in zip(
-            model_configs, expected_model_configs
-        ):
-            self.assertEqual(expected_model_config.cpu_only(), model_config.cpu_only())
-            self.assertEqual(
-                expected_model_config.model_name(), model_config.model_name()
-            )
-            self.assertEqual(
-                expected_model_config.parameters(), model_config.parameters()
-            )
-            self.assertEqual(
-                expected_model_config.constraints(), model_config.constraints()
-            )
-            self.assertEqual(
-                expected_model_config.objectives(), model_config.objectives()
-            )
-            self.assertEqual(
-                expected_model_config.model_config_parameters(),
-                model_config.model_config_parameters(),
-            )
-
-    def _assert_equality_of_plot_configs(self, plot_configs, expected_plot_configs):
-        self.assertEqual(len(plot_configs), len(expected_plot_configs))
-        for plot_config, expected_plot_config in zip(
-            plot_configs, expected_plot_configs
-        ):
-            self.assertEqual(expected_plot_config.name(), plot_config.name())
-            self.assertEqual(expected_plot_config.title(), plot_config.title())
-            self.assertEqual(expected_plot_config.x_axis(), plot_config.x_axis())
-            self.assertEqual(expected_plot_config.y_axis(), plot_config.y_axis())
-            self.assertEqual(expected_plot_config.monotonic(), plot_config.monotonic())
-
-    def _assert_model_config_types(self, model_config):
-        self.assertIsInstance(model_config.field_type(), ConfigUnion)
-
-        if isinstance(model_config.field_type().raw_value(), ConfigListGeneric):
-            self.assertIsInstance(
-                model_config.field_type().raw_value().container_type(), ConfigUnion
-            )
-        else:
-            self.assertIsInstance(model_config.field_type().raw_value(), ConfigObject)
-
-    def _assert_model_object_types(
-        self,
-        model_config,
-        model_name,
-        check_parameters=False,
-        check_concurrency=False,
-        check_batch_size=False,
-    ):
-        if isinstance(model_config, ConfigUnion):
-            self.assertIsInstance(model_config.raw_value(), ConfigObject)
-            self.assertIsInstance(
-                model_config.raw_value().raw_value()[model_name], ConfigObject
-            )
-        else:
-            self.assertIsInstance(model_config, ConfigObject)
-            if check_parameters:
-                parameters_config = model_config.raw_value()["parameters"]
-                self.assertIsInstance(parameters_config, ConfigObject)
-
-                if check_concurrency:
-                    self.assertIsInstance(
-                        parameters_config.raw_value()["concurrency"], ConfigListNumeric
-                    )
-
-    def _assert_model_config_params(self, model_config_parameters):
-        self.assertIsInstance(model_config_parameters, ConfigObject)
-
-        input_param = model_config_parameters.raw_value()["input"]
-        self.assertIsInstance(input_param, ConfigSweep)
-        # Is list of params
-        self.assertIsInstance(input_param.raw_value(), ConfigListGeneric)
-
-        # Each subitem is also a list
-        self.assertIsInstance(input_param.raw_value().container_type(), ConfigUnion)
-
-        single_sweep_param = input_param.raw_value().raw_value()[0].raw_value()
-        self.assertIsInstance(single_sweep_param.raw_value()[0], ConfigObject)
-
-        # Check types for 'name'
-        name_param = single_sweep_param.raw_value()[0].raw_value()["name"]
-        self.assertIsInstance(name_param, ConfigSweep)
-        self.assertIsInstance(name_param.raw_value(), ConfigListGeneric)
-        self.assertIsInstance(name_param.raw_value().container_type(), ConfigUnion)
-
-        # Check types for 'data_type'
-        data_type_param = single_sweep_param.raw_value()[0].raw_value()["data_type"]
-        self.assertIsInstance(data_type_param, ConfigSweep)
-        self.assertIsInstance(data_type_param.raw_value(), ConfigListGeneric)
-        self.assertIsInstance(data_type_param.raw_value().container_type(), ConfigUnion)
-        self.assertIsInstance(data_type_param.raw_value().raw_value()[0], ConfigUnion)
-        self.assertIsInstance(
-            data_type_param.raw_value().raw_value()[0].raw_value(), ConfigEnum
-        )
-
-        # Check types for 'dims'
-        dims_param = single_sweep_param.raw_value()[0].raw_value()["dims"]
-        self.assertIsInstance(dims_param, ConfigSweep)
-        self.assertIsInstance(dims_param.raw_value(), ConfigListGeneric)
-        self.assertIsInstance(dims_param.raw_value().container_type(), ConfigUnion)
-        self.assertIsInstance(dims_param.raw_value().raw_value()[0], ConfigUnion)
-        self.assertIsInstance(
-            dims_param.raw_value().raw_value()[0].raw_value(), ConfigListGeneric
-        )
-        self.assertIsInstance(
-            dims_param.raw_value().raw_value()[0].raw_value().raw_value()[0],
-            ConfigPrimitive,
-        )
-
-        # Check types for 'format'
-        format_param = single_sweep_param.raw_value()[0].raw_value()["format"]
-        self.assertIsInstance(format_param, ConfigSweep)
-        self.assertIsInstance(format_param.raw_value(), ConfigListGeneric)
-        self.assertIsInstance(format_param.raw_value().container_type(), ConfigUnion)
-
-    def _assert_model_str_type(self, model_config):
-        self.assertIsInstance(model_config, ConfigUnion)
-        self.assertIsInstance(model_config.raw_value(), ConfigPrimitive)
-
-    def setUp(self):
-        # Mock path validation
-        self.mock_os = MockOSMethods(
-            mock_paths=["model_analyzer.config.input.config_utils"]
-        )
-        self.mock_os.start()
-
-    def tearDown(self):
-        self.mock_os.stop()
-        patch.stopall()
-
-    def test_config(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-            "--profile-models",
-            "vgg11",
-        ]
-        yaml_content = "model_repository: yaml_repository"
-        config = self._evaluate_config(args, yaml_content)
-
-        # CLI flag has the highest priority
-        self.assertEqual(config.get_all_config()["model_repository"], "cli_repository")
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "-f",
-            "path-to-config-file",
-            "--profile-models",
-            "vgg11",
-        ]
-        yaml_content = "model_repository: yaml_repository"
-        config = self._evaluate_config(args, yaml_content)
-
-        # If CLI flag doesn't exist, YAML config has the highest priority
-        self.assertEqual(config.get_all_config()["model_repository"], "yaml_repository")
-
-        args = ["model-analyzer", "profile", "-f", "path-to-config-file"]
-        yaml_content = "model_repository: yaml_repository"
-        self._assert_error_on_evaluate_config(args, yaml_content)
-
-    def test_missing_config(self):
-        """Test that we fail if the required option profile-models is omitted"""
-
-        args = ["model-analyzer", "profile", "-f", "path-to-config-file"]
-        yaml_content = "model_repository: yaml_repository"
-        self._assert_error_on_evaluate_config(args, yaml_content)
-
-    def test_conflicting_configs(self):
-        """Test that we fail if an option is specified in both CLI and YAML"""
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-            "--profile-models",
-            "vgg11",
-        ]
-        yaml_content = "profile_models: add_sub"
-        self._assert_error_on_evaluate_config(args, yaml_content)
-
-    def test_range_and_list_values(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = "profile_models: model_1,model_2"
-        config = self._evaluate_config(args, yaml_content)
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "model_1",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-            ),
-            ConfigModelProfileSpec(
-                "model_2",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-            ),
-        ]
-        self._assert_equality_of_model_configs(
-            config.get_all_config()["profile_models"], expected_model_configs
-        )
-        self.assertIsInstance(
-            config.get_config()["profile_models"].field_type().raw_value(), ConfigObject
-        )
-
-        yaml_content = """
-profile_models:
-    - model_1
-    - model_2
-"""
-        config = self._evaluate_config(args, yaml_content)
-        self._assert_equality_of_model_configs(
-            config.get_all_config()["profile_models"], expected_model_configs
-        )
-
-        model_config = config.get_config()["profile_models"]
-
-        self._assert_model_config_types(model_config)
-        self.assertIsInstance(
-            model_config.field_type()
-            .raw_value()
-            .raw_value()["model_1"]
-            .raw_value()["objectives"],
-            ConfigUnion,
-        )
-        self.assertIsInstance(
-            model_config.field_type()
-            .raw_value()
-            .raw_value()["model_1"]
-            .raw_value()["parameters"],
-            ConfigObject,
-        )
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-            "--profile-models",
-            "model_1,model_2",
-        ]
-        yaml_content = """
-batch_sizes:
-    - 2
-    - 3
-"""
-        config = self._evaluate_config(args, yaml_content)
-        self.assertEqual(config.get_all_config()["batch_sizes"], [2, 3])
-        self.assertIsInstance(
-            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
-        )
-
-        yaml_content = """
-batch_sizes: 2
-"""
-        config = self._evaluate_config(args, yaml_content)
-        self.assertEqual(config.get_all_config()["batch_sizes"], [2])
-        self.assertIsInstance(
-            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
-        )
-
-        yaml_content = """
-concurrency: 2
-"""
-        config = self._evaluate_config(args, yaml_content)
-        self.assertEqual(config.get_all_config()["concurrency"], [2])
-        self.assertIsInstance(
-            config.get_config()["concurrency"].field_type(), ConfigListNumeric
-        )
-        self.assertEqual(config.get_all_config()["batch_sizes"], [1])
-        self.assertIsInstance(
-            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
-        )
-
-        yaml_content = """
-batch_sizes:
-    start: 2
-    stop: 6
-"""
-        config = self._evaluate_config(args, yaml_content)
-        self.assertEqual(config.get_all_config()["batch_sizes"], [2, 3, 4, 5, 6])
-        self.assertIsInstance(
-            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
-        )
-
-        yaml_content = """
-batch_sizes:
-    start: 2
-    stop: 6
-    step: 2
-"""
-        config = self._evaluate_config(args, yaml_content)
-        self.assertEqual(config.get_all_config()["batch_sizes"], [2, 4, 6])
-        self.assertIsInstance(
-            config.get_config()["batch_sizes"].field_type(), ConfigListNumeric
-        )
-
-    def test_object(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-profile_models:
-  -
-    vgg_16_graphdef:
-      parameters:
-        concurrency:
-          - 1
-          - 2
-          - 3
-          - 4
-  - vgg_19_graphdef
-"""
-        config = self._evaluate_config(args, yaml_content)
-        model_config = config.get_config()["profile_models"]
-        self._assert_model_config_types(model_config)
-
-        expected_model_objects = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={
-                    "batch_sizes": [1],
-                    "concurrency": [1, 2, 3, 4],
-                    "request_rate": [],
-                },
-                objectives={"perf_throughput": 10},
-            ),
-            ConfigModelProfileSpec(
-                "vgg_19_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-            ),
-        ]
-
-        # Check the types for the first value
-        first_model = (
-            model_config.field_type().raw_value().raw_value()["vgg_16_graphdef"]
-        )
-        self._assert_model_object_types(
-            first_model,
-            "vgg_16_graphdef",
-            check_parameters=True,
-            check_concurrency=True,
-        )
-
-        # Check the types for the second value
-        second_model = (
-            model_config.field_type().raw_value().raw_value()["vgg_19_graphdef"]
-        )
-        self._assert_model_object_types(second_model, "vgg_19_graphdef")
-
-        self._assert_equality_of_model_configs(
-            config.get_all_config()["profile_models"], expected_model_objects
-        )
-
-        yaml_content = """
-profile_models:
-  vgg_16_graphdef:
-    parameters:
-      concurrency:
-        - 1
-        - 2
-        - 3
-        - 4
-  vgg_19_graphdef:
-    parameters:
-      concurrency:
-        - 1
-        - 2
-        - 3
-        - 4
-      batch_sizes:
-          start: 2
-          stop: 6
-          step: 2
-"""
-        expected_model_objects = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={
-                    "batch_sizes": [1],
-                    "concurrency": [1, 2, 3, 4],
-                    "request_rate": [],
-                },
-                objectives={"perf_throughput": 10},
-            ),
-            ConfigModelProfileSpec(
-                "vgg_19_graphdef",
-                parameters={
-                    "concurrency": [1, 2, 3, 4],
-                    "batch_sizes": [2, 4, 6],
-                    "request_rate": [],
-                },
-                objectives={"perf_throughput": 10},
-            ),
-        ]
-        config = self._evaluate_config(args, yaml_content)
-        self._assert_equality_of_model_configs(
-            config.get_all_config()["profile_models"], expected_model_objects
-        )
-
-        model_config = config.get_config()["profile_models"]
-        self._assert_model_config_types(model_config)
-
-        # first model
-        first_model = (
-            model_config.field_type().raw_value().raw_value()["vgg_16_graphdef"]
-        )
-        self._assert_model_object_types(
-            first_model,
-            "vgg_16_graphdef",
-            check_parameters=True,
-            check_concurrency=True,
-            check_batch_size=True,
-        )
-
-        # second model
-        second_model = (
-            model_config.field_type().raw_value().raw_value()["vgg_19_graphdef"]
-        )
-        self._assert_model_object_types(
-            second_model,
-            "vgg_19_graphdef",
-            check_parameters=True,
-            check_concurrency=True,
-            check_batch_size=True,
-        )
-
-    def test_constraints(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-profile_models:
-  -
-    vgg_16_graphdef:
-      parameters:
-        concurrency:
-          - 1
-          - 2
-          - 3
-          - 4
-      objectives:
-        perf_throughput: 10
-        gpu_used_memory: 5
-      constraints:
-        gpu_used_memory:
-          max: 80
-  - vgg_19_graphdef
-"""
-        config = self._evaluate_config(args, yaml_content)
-        expected_model_objects = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={
-                    "batch_sizes": [1],
-                    "concurrency": [1, 2, 3, 4],
-                    "request_rate": [],
-                },
-                objectives={"perf_throughput": 10, "gpu_used_memory": 5},
-                constraints={
-                    "gpu_used_memory": {
-                        "max": 80,
-                    }
-                },
-            ),
-            ConfigModelProfileSpec(
-                "vgg_19_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-            ),
-        ]
-        self._assert_equality_of_model_configs(
-            config.get_all_config()["profile_models"], expected_model_objects
-        )
-
-        # GPU Memory shouldn't have min
-        yaml_content = """
-profile_models:
-  -
-    vgg_16_graphdef:
-      parameters:
-        concurrency:
-          - 1
-          - 2
-          - 3
-          - 4
-      objectives:
-        - perf_throughput
-        - gpu_used_memory
-      constraints:
-        gpu_memory:
-          max: 80
-          min: 45
-  - vgg_19_graphdef
-"""
-        self._assert_error_on_evaluate_config(args, yaml_content)
-
-        # Test objective key that is not one of the supported metrics
-        yaml_content = """
-profile_models:
-  -
-    vgg_16_graphdef:
-      parameters:
-        concurrency:
-          - 1
-          - 2
-          - 3
-          - 4
-      objectives:
-        - throughput
-      constraints:
-        gpu_used_memory:
-          max: 80
-  - vgg_19_graphdef
-"""
-        self._assert_error_on_evaluate_config(args, yaml_content)
-
-    def test_validation(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-
-        # end key should not be included in concurrency
-        yaml_content = """
-profile_models:
-  -
-    vgg_16_graphdef:
-      parameters:
-        concurrency:
-            start: 4
-            stop: 12
-            end: 2
-"""
-        self._assert_error_on_evaluate_config(args, yaml_content)
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-
-        yaml_content = """
-profile_models:
-  -
-    vgg_16_graphdef:
-      parameters:
-        concurrency:
-            start: 13
-            stop: 12
-"""
-        self._assert_error_on_evaluate_config(args, yaml_content)
-
-    def test_config_model(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-profile_models:
-  -
-    vgg_16_graphdef:
-        model_config_parameters:
-            instance_group:
-                -
-                    kind: KIND_GPU
-                    count: 1
-
-"""
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-                model_config_parameters={
-                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        yaml_content = """
-profile_models:
-  -
-    vgg_16_graphdef:
-        model_config_parameters:
-            instance_group:
-                -
-                    kind: KIND_GPU
-                    count: 1
-
-"""
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-                model_config_parameters={
-                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-            profile_models:
-            -
-                vgg_16_graphdef:
-                    model_config_parameters:
-                        instance_group:
-                            -
-                                kind: KIND_GPU
-                                count: 1
-                            -
-                                kind: KIND_CPU
-                                count: 1
-
-            """
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-                model_config_parameters={
-                    "instance_group": [
-                        [
-                            {"kind": ["KIND_GPU"], "count": [1]},
-                            {"kind": ["KIND_CPU"], "count": [1]},
-                        ]
-                    ]
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-            profile_models:
-            -
-                vgg_16_graphdef:
-                    model_config_parameters:
-                        instance_group:
-                        -
-                            -
-                                kind: KIND_GPU
-                                count: 1
-                        -
-                            -
-                                kind: KIND_CPU
-                                count: 1
-
-            """
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-                model_config_parameters={
-                    "instance_group": [
-                        [{"kind": ["KIND_GPU"], "count": [1]}],
-                        [{"kind": ["KIND_CPU"], "count": [1]}],
-                    ]
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        yaml_content = """
-            profile_models:
-            -
-                vgg_16_graphdef:
-                    model_config_parameters:
-                        input:
-                        -
-                            name: NV_MODEL_INPUT
-                            data_type: TYPE_FP32
-                            format: FORMAT_NHWC
-                            dims: [256, 256, 3]
-
-            """
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-                model_config_parameters={
-                    "input": [
-                        [
-                            {
-                                "name": ["NV_MODEL_INPUT"],
-                                "data_type": ["TYPE_FP32"],
-                                "format": ["FORMAT_NHWC"],
-                                "dims": [[256, 256, 3]],
-                            }
-                        ]
-                    ]
-                },
-            )
-        ]
-
-        model_config = config.get_config()["profile_models"]
-        self._assert_model_config_types(model_config)
-
-        model = model_config.field_type().raw_value().raw_value()["vgg_16_graphdef"]
-        self._assert_model_object_types(model, "vgg_16_graphdef")
-
-        model_config_parameters = model.raw_value()["model_config_parameters"]
-
-        self._assert_model_config_params(model_config_parameters)
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        yaml_content = """
-            profile_models:
-            -
-                vgg_16_graphdef:
-                    perf_analyzer_flags:
-                        measurement-interval: 10000
-                        model-version: 2
-                        streaming: "header:value"
-
-            """
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-                perf_analyzer_flags={
-                    "measurement-interval": 10000,
-                    "model-version": 2,
-                    "streaming": "header:value",
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        yaml_content = """
-            profile_models:
-            -
-                vgg_16_graphdef:
-                    perf_analyzer_flags:
-                        measurement-interval: 10000
-                        model-version: 2
-                        shape: ["name1:1,2,3", "name2:4,5,6"]
-
-            """
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-                perf_analyzer_flags={
-                    "measurement-interval": 10000,
-                    "model-version": 2,
-                    "shape": ["name1:1,2,3", "name2:4,5,6"],
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        yaml_content = """
-            profile_models:
-            -
-                vgg_16_graphdef:
-                    perf_analyzer_flags:
-                        latency_report_file: ["file1", "file2"]
-
-            """
-        with self.assertRaises(TritonModelAnalyzerException):
-            # latency_report_file is not additive
-            self._evaluate_config(args, yaml_content)
-
-        yaml_content = """
-            profile_models:
-            -
-                vgg_16_graphdef:
-                    perf_analyzer_flags:
-                        disallowed-perf-flag: some_value
-
-            """
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content)
-
-    def test_config_sweep(self):
-        config_sweep = ConfigSweep(ConfigPrimitive(int))
-        config_sweep.set_value(2)
-
-    def test_config_plot(self):
-        args = ["model-analyzer", "report", "-f", "path-to-config-file"]
-        yaml_content = """
-            report_model_configs: vgg_16_graphdef
-            plots:
-                test_plot:
-                    title: Throughput vs. Latency
-                    x_axis: perf_throughput
-                    y_axis: perf_latency_p99
-            """
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        plot_configs = config.get_all_config()["plots"]
-        expected_plot_configs = [
-            ConfigPlot(
-                "test_plot",
-                title="Throughput vs. Latency",
-                x_axis="perf_throughput",
-                y_axis="perf_latency_p99",
-            )
-        ]
-        self._assert_equality_of_plot_configs(plot_configs, expected_plot_configs)
-
-        yaml_content = """
-            report_model_configs: vgg_16_graphdef
-            plots:
-            - test_plot1:
-                title: Throughput vs. Latency
-                x_axis: perf_throughput
-                y_axis: perf_latency_p99
-            - test_plot2:
-                title: GPU Memory vs. Latency
-                x_axis: gpu_used_memory
-                y_axis: perf_latency_p99
-            """
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        plot_configs = config.get_all_config()["plots"]
-        expected_plot_configs = [
-            ConfigPlot(
-                "test_plot1",
-                title="Throughput vs. Latency",
-                x_axis="perf_throughput",
-                y_axis="perf_latency_p99",
-            ),
-            ConfigPlot(
-                "test_plot2",
-                title="GPU Memory vs. Latency",
-                x_axis="gpu_used_memory",
-                y_axis="perf_latency_p99",
-            ),
-        ]
-        self._assert_equality_of_plot_configs(plot_configs, expected_plot_configs)
-
-        yaml_content = """
-            report_model_configs: vgg_16_graphdef
-            plots:
-                test_plot1:
-                    title: Throughput vs. Latency
-                    x_axis: perf_throughput
-                    y_axis: perf_latency_p99
-                test_plot2:
-                    title: GPU Memory vs. Latency
-                    x_axis: gpu_used_memory
-                    y_axis: perf_latency_p99
-            """
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        plot_configs = config.get_all_config()["plots"]
-        expected_plot_configs = [
-            ConfigPlot(
-                "test_plot1",
-                title="Throughput vs. Latency",
-                x_axis="perf_throughput",
-                y_axis="perf_latency_p99",
-            ),
-            ConfigPlot(
-                "test_plot2",
-                title="GPU Memory vs. Latency",
-                x_axis="gpu_used_memory",
-                y_axis="perf_latency_p99",
-            ),
-        ]
-        self._assert_equality_of_plot_configs(plot_configs, expected_plot_configs)
-
-        yaml_content = """
-            report_model_configs:
-              vgg_16_graphdef:
-                plots:
-                    test_plot1:
-                        title: Throughput vs. Latency
-                        x_axis: perf_throughput
-                        y_axis: perf_latency_p99
-                        monotonic: True
-                    test_plot2:
-                        title: GPU Memory vs. Latency
-                        x_axis: gpu_used_memory
-                        y_axis: perf_latency_p99
-                        monotonic: False
-            """
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        plot_configs = config.get_all_config()["report_model_configs"][0].plots()
-        expected_plot_configs = [
-            ConfigPlot(
-                "test_plot1",
-                title="Throughput vs. Latency",
-                x_axis="perf_throughput",
-                y_axis="perf_latency_p99",
-                monotonic=True,
-            ),
-            ConfigPlot(
-                "test_plot2",
-                title="GPU Memory vs. Latency",
-                x_axis="gpu_used_memory",
-                y_axis="perf_latency_p99",
-                monotonic=False,
-            ),
-        ]
-        self._assert_equality_of_plot_configs(plot_configs, expected_plot_configs)
-
-    def test_error_messages(self):
-        # ConfigListNumeric
-        config_numeric = ConfigListNumeric(float)
-        config_numeric.set_name("key")
-        config_status = config_numeric.set_value(
-            {"start": 12, "stop": 15, "undefined_key": 8}
-        )
-
-        config_message = config_status.message()
-        result = re.search(".*'start'.*'stop'.*'undefined_key'.*", config_message)
-
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-        self.assertIsNotNone(config_message)
-        self.assertIsNotNone(result)
-
-        config_numeric = ConfigListNumeric(float)
-        config_numeric.set_name("key")
-        config_status = config_numeric.set_value({"start": 12, "stop": "two"})
-        print(config_message)
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        config_numeric = ConfigListNumeric(float)
-        config_numeric.set_name("key")
-        config_status = config_numeric.set_value({"start": "five", "stop": 2})
-        config_message = config_status.message()
-        print(config_message)
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        config_numeric = ConfigListNumeric(float)
-        config_numeric.set_name("key")
-        config_status = config_numeric.set_value({"start": 10, "stop": 2})
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        # ConfigUnion error message
-        config_union = ConfigUnion([ConfigListNumeric(float), ConfigPrimitive(str)])
-        config_union.set_name("key")
-
-        # Dictionaries are not accepted.
-        config_status = config_union.set_value({"a": "b"})
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        # ConfigEnum
-        config_enum = ConfigEnum(["a", "b"])
-        config_enum.set_name("key")
-        config_status = config_enum.set_value("c")
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        # ConfigListGeneric
-        config_list_generic = ConfigListGeneric(ConfigPrimitive(float))
-        config_list_generic.set_name("key")
-        config_status = config_list_generic.set_value({"a": "b"})
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        # ConfigListString
-        config_list_string = ConfigListString()
-        config_list_string.set_name("key")
-        config_status = config_list_string.set_value({"a": "b"})
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        config_status = config_list_string.set_value([{"a": "b"}])
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        # ConfigObject
-        config_object = ConfigObject(schema={"key": ConfigPrimitive(float)})
-        config_object.set_name("key")
-        config_status = config_object.set_value({"undefiend_key": 2.0})
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        config_status = config_object.set_value({"key": [1, 2, 3]})
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        config_status = config_object.set_value([1, 2, 3])
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-        # ConfigPrimitive
-        config_primitive = ConfigPrimitive(float)
-        config_primitive.set_name("key")
-        config_status = config_primitive.set_value("a")
-        print(config_status.message())
-        self.assertEqual(config_status.status(), CONFIG_PARSER_FAILURE)
-
-    def test_autofill(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-        profile_models:
-        -
-            vgg_16_graphdef:
-                model_config_parameters:
-                    instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: 1
-
-        """
-        # Test defaults
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={"batch_sizes": [1], "concurrency": [], "request_rate": []},
-                objectives={"perf_throughput": 10},
-                model_config_parameters={
-                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        yaml_content = """
-objectives:
-  perf_throughput: 10
-  gpu_used_memory: 5
-constraints:
-  gpu_used_memory:
-    max: 80
-profile_models:
-  -
-    vgg_16_graphdef:
-        parameters:
-          batch_sizes:
-            - 16
-            - 32
-          concurrency:
-            start: 2
-            stop: 4
-            step: 2
-        model_config_parameters:
-            instance_group:
-                -
-                    kind: KIND_GPU
-                    count: 1
-
-"""
-        # Test autofill objectives and constraints
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={
-                    "batch_sizes": [16, 32],
-                    "concurrency": [2, 4],
-                    "request_rate": [],
-                },
-                objectives={"perf_throughput": 10, "gpu_used_memory": 5},
-                constraints={
-                    "gpu_used_memory": {
-                        "max": 80,
-                    }
-                },
-                model_config_parameters={
-                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        yaml_content = """
-concurrency:
-  start: 2
-  stop : 4
-  step: 2
-profile_models:
-  -
-    vgg_16_graphdef:
-        parameters:
-          batch_sizes:
-            - 16
-            - 32
-        objectives:
-          gpu_used_memory: 10
-        constraints:
-          perf_latency_p99:
-            max: 8000
-        model_config_parameters:
-            instance_group:
-                -
-                    kind: KIND_GPU
-                    count: 1
-
-"""
-        # Test autofill concurrency
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={
-                    "batch_sizes": [16, 32],
-                    "concurrency": [2, 4],
-                    "request_rate": [],
-                },
-                objectives={"gpu_used_memory": 10},
-                constraints={"perf_latency_p99": {"max": 8000}},
-                model_config_parameters={
-                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        yaml_content = """
-batch_sizes:
-  - 16
-  - 32
-profile_models:
-  -
-    vgg_16_graphdef:
-        parameters:
-          concurrency:
-            start: 2
-            stop : 4
-            step: 2
-        objectives:
-          - gpu_used_memory
-        constraints:
-          perf_latency_p99:
-            max: 8000
-        model_config_parameters:
-            instance_group:
-                -
-                    kind: KIND_GPU
-                    count: 1
-
-"""
-        # Test autofill batch sizes
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={
-                    "batch_sizes": [16, 32],
-                    "concurrency": [2, 4],
-                    "request_rate": [],
-                },
-                objectives={"gpu_used_memory": 10},
-                constraints={"perf_latency_p99": {"max": 8000}},
-                model_config_parameters={
-                    "instance_group": [[{"kind": ["KIND_GPU"], "count": [1]}]]
-                },
-            )
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        yaml_content = """
-objectives:
-  perf_throughput: 10
-  perf_latency_p99: 5
-constraints:
-    perf_latency_p99:
-      max: 8000
-    gpu_used_memory:
-      max: 10000
-batch_sizes:
-  - 16
-  - 32
-concurrency:
-  start: 2
-  stop: 4
-  step: 2
-profile_models:
-  -
-    vgg_16_graphdef:
-        parameters:
-          concurrency:
-            start: 5
-            stop : 7
-        objectives:
-          - gpu_used_memory
-  -
-    vgg_19_graphdef:
-        parameters:
-          batch_sizes:
-            - 1
-            - 2
-        constraints:
-          perf_latency_p99:
-            max: 8000
-"""
-        # Test autofill batch sizes
-        config = self._evaluate_config(args, yaml_content)
-        model_configs = config.get_all_config()["profile_models"]
-        expected_model_configs = [
-            ConfigModelProfileSpec(
-                "vgg_16_graphdef",
-                parameters={
-                    "batch_sizes": [16, 32],
-                    "concurrency": [5, 6, 7],
-                    "request_rate": [],
-                },
-                objectives={"gpu_used_memory": 10},
-                constraints={
-                    "perf_latency_p99": {"max": 8000},
-                    "gpu_used_memory": {"max": 10000},
-                },
-            ),
-            ConfigModelProfileSpec(
-                "vgg_19_graphdef",
-                parameters={
-                    "batch_sizes": [1, 2],
-                    "concurrency": [2, 4],
-                    "request_rate": [],
-                },
-                objectives={"perf_throughput": 10, "perf_latency_p99": 5},
-                constraints={"perf_latency_p99": {"max": 8000}},
-            ),
-        ]
-        self._assert_equality_of_model_configs(model_configs, expected_model_configs)
-
-        # Test autofill CPU_ONLY. It will only be false if no local gpus are available AND we are not in remote mode
-        yaml_content = """
-profile_models:
-  - vgg_16_graphdef
-"""
-        for launch_mode in ["remote", "c_api", "docker", "local"]:
-            for local_gpus_available in [True, False]:
-                new_args = args.copy()
-                new_args.extend(["--triton-launch-mode", launch_mode])
-                config = self._evaluate_config(
-                    new_args, yaml_content, numba_available=local_gpus_available
-                )
-                model_configs = config.get_all_config()["profile_models"]
-                expected_cpu_only = not local_gpus_available and launch_mode != "remote"
-                expected_model_configs = [
-                    ConfigModelProfileSpec(
-                        "vgg_16_graphdef",
-                        cpu_only=expected_cpu_only,
-                        parameters=self._create_parameters(batch_sizes=[1]),
-                        objectives={"perf_throughput": 10},
-                    )
-                ]
-                self._assert_equality_of_model_configs(
-                    model_configs, expected_model_configs
-                )
-
-    def test_config_shorthands(self):
-        """
-        test flags like --latency-budget
-        """
-
-        for constraint_shorthand in [
-            ("--latency-budget", "max", "perf_latency_p99"),
-            ("--min-throughput", "min", "perf_throughput"),
-        ]:
-            args = [
-                "model-analyzer",
-                "profile",
-                "--profile-models",
-                "test_model",
-                constraint_shorthand[0],
-                "40",
-                "--model-repository",
-                "cli_repository",
-            ]
-            # check that global and model specific constraints are filled
-            yaml_content = ""
-            config = self._evaluate_config(args, yaml_content, subcommand="profile")
-            self.assertDictEqual(
-                config.get_all_config()["constraints"],
-                {constraint_shorthand[2]: {constraint_shorthand[1]: 40}},
-            )
-
-            self.assertEqual(
-                config.get_all_config()["profile_models"][0].constraints(),
-                ModelConstraints(
-                    {constraint_shorthand[2]: {constraint_shorthand[1]: 40}}
-                ),
-            )
-
-            # check that model specific constraints are appended to
-            args = [
-                "model-analyzer",
-                "profile",
-                constraint_shorthand[0],
-                "40",
-                "-f",
-                "path-to-config-file",
-                "--model-repository",
-                "cli_repository",
-            ]
-            yaml_content = """
-            profile_models:
-                test_model:
-                    constraints:
-                        gpu_used_memory:
-                            max : 100
-            """
-            config = self._evaluate_config(args, yaml_content, subcommand="profile")
-            self.assertDictEqual(
-                config.get_all_config()["constraints"],
-                {constraint_shorthand[2]: {constraint_shorthand[1]: 40}},
-            )
-            self.assertEqual(
-                config.get_all_config()["profile_models"][0].constraints(),
-                ModelConstraints(
-                    {
-                        constraint_shorthand[2]: {constraint_shorthand[1]: 40},
-                        "gpu_used_memory": {"max": 100},
-                    }
-                ),
-            )
-
-            # check that model specific constraints are replaced
-            yaml_content = f"""
-            profile_models:
-                test_model:
-                    constraints:
-                        {constraint_shorthand[2]}:
-                            {constraint_shorthand[1]} : 100
-            """
-            config = self._evaluate_config(args, yaml_content, subcommand="profile")
-            self.assertEqual(
-                config.get_all_config()["profile_models"][0].constraints(),
-                ModelConstraints(
-                    {constraint_shorthand[2]: {constraint_shorthand[1]: 40}}
-                ),
-            )
-
-            # check that global constraints are appended to
-            yaml_content = """
-            profile_models: test_model
-            constraints:
-                gpu_used_memory:
-                    max : 100
-            """
-
-            config = self._evaluate_config(args, yaml_content, subcommand="profile")
-            self.assertDictEqual(
-                config.get_all_config()["constraints"],
-                {
-                    constraint_shorthand[2]: {constraint_shorthand[1]: 40},
-                    "gpu_used_memory": {"max": 100},
-                },
-            )
-
-            # check that global constraints are replaced
-            yaml_content = f"""
-            profile_models: test_model
-            constraints:
-                {constraint_shorthand[2]}:
-                    {constraint_shorthand[1]} : 100
-            """
-            config = self._evaluate_config(args, yaml_content, subcommand="profile")
-            self.assertDictEqual(
-                config.get_all_config()["constraints"],
-                {constraint_shorthand[2]: {constraint_shorthand[1]: 40}},
-            )
-
-    def test_triton_server_flags(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-            profile_models: model1, model2
-            triton_server_flags:
-                strict-model-config: false
-                backend-config: test_backend_config
-            """
-        config = self._evaluate_config(args, yaml_content)
-        self.assertDictEqual(
-            config.get_all_config()["triton_server_flags"],
-            {"strict-model-config": "False", "backend-config": "test_backend_config"},
-        )
-
-        yaml_content = """
-            profile_models: model1, model2
-            triton_server_flags:
-                disallowed-config-option: some_value
-                backend-config: test_backend_config
-            """
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content)
-
-        yaml_content = """
-            profile_models:
-                model1:
-                    triton_server_flags:
-                        strict_model_config: false
-                        backend_config: test_backend_config
-            """
-        config = self._evaluate_config(args, yaml_content)
-        self.assertDictEqual(
-            config.get_all_config()["profile_models"][0].triton_server_flags(),
-            {"strict_model_config": "False", "backend_config": "test_backend_config"},
-        )
-
-    def test_triton_server_environment(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-            profile_models: model1, model2
-            triton_server_environment:
-                LD_PRELOAD: libtest.so
-                LD_LIBRARY_PATH: /path/to/test/lib
-            """
-        config = self._evaluate_config(args, yaml_content)
-        self.assertDictEqual(
-            config.get_all_config()["triton_server_environment"],
-            {"LD_PRELOAD": "libtest.so", "LD_LIBRARY_PATH": "/path/to/test/lib"},
-        )
-
-        yaml_content = """
-            profile_models:
-                model1:
-                    triton_server_environment:
-                        LD_PRELOAD: libtest.so
-                        LD_LIBRARY_PATH: /path/to/test/lib
-            """
-        config = self._evaluate_config(args, yaml_content)
-        self.assertDictEqual(
-            config.get_all_config()["profile_models"][0].triton_server_environment(),
-            {"LD_PRELOAD": "libtest.so", "LD_LIBRARY_PATH": "/path/to/test/lib"},
-        )
-
-    def test_triton_docker_args(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-            "--triton-launch-mode",
-            "docker",
-        ]
-        yaml_content = """
-            profile_models: model1, model2
-            triton_docker_args:
-                image: foo
-                command: goo
-            """
-        config = self._evaluate_config(args, yaml_content)
-        self.assertDictEqual(
-            config.get_all_config()["triton_docker_args"],
-            {"image": "foo", "command": "goo"},
-        )
-
-        yaml_content = """
-            profile_models:
-                model1:
-                    triton_docker_args:
-                        image: foo
-                        command: goo
-            """
-        config = self._evaluate_config(args, yaml_content)
-        self.assertDictEqual(
-            config.get_all_config()["profile_models"][0].triton_docker_args(),
-            {"image": "foo", "command": "goo"},
-        )
-
-        # This is testing that all options can be parsed by the YAML
-        yaml_content = """
-            profile_models: model1, model2
-            triton_docker_args:
-                image: abc
-                command: abc
-                auto_remove: true
-                blkio_weight_device: ['abc']
-                blkio_weight: 1
-                cap_add: ['abc']
-                cap_drop: ['abc']
-                cgroup_parent: abc
-                cgroupns: abc
-                cpu_count: 1
-                cpu_percent: 1
-                cpu_period: 1
-                cpu_quota: 1
-                cpu_rt_period: 1
-                cpu_shares: 1
-                cpuset_cpus: abc
-                cpuset_mems: abc
-                detach: true
-                domainname: abc
-                entrypoint: abc
-                environment: ['abc']
-                hostname: abc
-                init: true
-                init_path: abc
-                ipc_mode: abc
-                isolation: abc
-                kernel_memory: abc
-                labels: abc
-                mac_address: abc
-                mem_limit: abc
-                mem_reservation: abc
-                memswap_limit: abc
-                name: abc
-                nano_cpus: 1
-                network: abc
-                network_disabled: true
-                network_mode: abc
-                oom_kill_disable: true
-                oom_score_adj: 1
-                pid_mode: abc
-                pids_limit: 1
-                platform: abc
-                privileged: true
-                publish_all_ports: true
-                remove: true
-                runtime: abc
-                shm_size: abc
-                stdin_open: true
-                stdout: true
-                stderr: true
-                stop_signal: abc
-                stream: true
-                tty: true
-                use_config_proxy: true
-                user: abc
-                userns_mode: abc
-                uts_mode: abc
-                version: abc
-                volume_driver: abc
-                volumes: ['abc']
-                working_dir: abc
-            """
-        _ = self._evaluate_config(args, yaml_content)
-
-    def test_report_configs(self):
-        args = ["model-analyzer", "report", "--report-model-configs", "test-model"]
-        yaml_content = ""
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        self.assertEqual(
-            config.get_all_config()["report_model_configs"][0].model_config_name(),
-            "test-model",
-        )
-
-        # check output format
-        args = [
-            "model-analyzer",
-            "report",
-            "--report-model-configs",
-            "test-model",
-            "--output-formats",
-            "pdf",
-        ]
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        self.assertEqual(config.get_all_config()["output_formats"], ["pdf"])
-        args = [
-            "model-analyzer",
-            "report",
-            "--report-model-configs",
-            "test-model",
-            "--output-formats",
-            "pdf,csv,svg",
-        ]
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        self.assertEqual(
-            config.get_all_config()["output_formats"], ["pdf", "csv", "svg"]
-        )
-
-        # Check yaml report model config
-        args = ["model-analyzer", "report", "-f", "path-to-config-file"]
-        yaml_content = """
-        report_model_configs:
-            - test_model_config_0
-        output_formats:
-          - pdf
-          - csv
-          - png
-        """
-
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        self.assertEqual(
-            config.get_all_config()["report_model_configs"][0].model_config_name(),
-            "test_model_config_0",
-        )
-        self.assertEqual(
-            config.get_all_config()["output_formats"], ["pdf", "csv", "png"]
-        )
-
-        # Check plots
-        args = ["model-analyzer", "report", "-f", "path-to-config-file"]
-        yaml_content = """
-        report_model_configs:
-           - test_model_config_0
-           - test_model_config_1
-        plots:
-            throughput_v_latency:
-                title: Throughput vs. Latency
-                x_axis: perf_latency_p99
-                y_axis: perf_throughput
-                monotonic: True
-        """
-
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        self.assertEqual(
-            config.get_all_config()["report_model_configs"][0].model_config_name(),
-            "test_model_config_0",
-        )
-        self.assertEqual(
-            config.get_all_config()["report_model_configs"][1].model_config_name(),
-            "test_model_config_1",
-        )
-        expected_config_plot = {
-            "throughput_v_latency": {
-                "title": "Throughput vs. Latency",
-                "x_axis": "perf_latency_p99",
-                "y_axis": "perf_throughput",
-                "monotonic": True,
-            }
-        }
-        config_plot = config.get_all_config()["plots"][0]
-        config_plot_dict = {
-            config_plot.name(): {
-                "title": config_plot.title(),
-                "x_axis": config_plot.x_axis(),
-                "y_axis": config_plot.y_axis(),
-                "monotonic": config_plot.monotonic(),
-            }
-        }
-        self.assertDictEqual(config_plot_dict, expected_config_plot)
-        for report_model_config in config.report_model_configs:
-            config_plot = report_model_config.plots()[0]
-            config_plot_dict = {
-                config_plot.name(): {
-                    "title": config_plot.title(),
-                    "x_axis": config_plot.x_axis(),
-                    "y_axis": config_plot.y_axis(),
-                    "monotonic": config_plot.monotonic(),
-                }
-            }
-            self.assertDictEqual(config_plot_dict, expected_config_plot)
-
-        # Check individual plots
-        yaml_content = """
-        report_model_configs:
-            test_model_config_0:
-                plots:
-                  model_specific_throughput_v_latency:
-                    title: model specific title
-                    x_axis: perf_latency_p99
-                    y_axis: perf_throughput
-                    monotonic: True
-        plots:
-            throughput_v_latency:
-                title: Throughput vs. Latency
-                x_axis: perf_latency_p99
-                y_axis: perf_throughput
-                monotonic: True
-        """
-
-        config = self._evaluate_config(args, yaml_content, subcommand="report")
-        global_config_plot = config.get_all_config()["plots"][0]
-        global_config_plot_dict = {
-            config_plot.name(): {
-                "title": global_config_plot.title(),
-                "x_axis": global_config_plot.x_axis(),
-                "y_axis": global_config_plot.y_axis(),
-                "monotonic": global_config_plot.monotonic(),
-            }
-        }
-        self.assertDictEqual(expected_config_plot, global_config_plot_dict)
-        model_specific_plot = config.get_all_config()["report_model_configs"][
-            0
-        ].plots()[0]
-        self.assertEqual(
-            model_specific_plot.name(), "model_specific_throughput_v_latency"
-        )
-        self.assertEqual(model_specific_plot.title(), "model specific title")
-
-    def test_path_validation(self):
-        # Test parent path validator
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "/",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-        checkpoint_directory: /test
-        profile_models:
-            - model1
-            - model2
-        """
-
-        self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        self.mock_os.set_os_path_exists_return_value(False)
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-        self.mock_os.set_os_path_exists_return_value(True)
-
-        # Test file path validator
-        yaml_content = """
-        triton_install_path: /opt/triton-model-analyzer/tests/test_config.py
-        profile_models:
-            - model1
-            - model2
-        """
-
-        self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        self.mock_os.set_os_path_exists_return_value(False)
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-        self.mock_os.set_os_path_exists_return_value(True)
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "-f",
-            "path-to-config-file",
-            "--model-repository",
-            "cli_repository",
-        ]
-        yaml_content = """
-        export_path: /opt/triton-model-analyzer/tests
-        profile_models:
-            - model1
-            - model2
-        """
-
-        self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        self.mock_os.set_os_path_exists_return_value(False)
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-        self.mock_os.set_os_path_exists_return_value(True)
-
-        # Test the binary path validator
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "/",
-            "-f",
-            "path-to-config-file",
-        ]
-        yaml_content = """
-        triton_server_path: tritonserver
-        profile_models:
-            - model1
-            - model2
-        """
-
-        self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        self.mock_os.set_os_path_exists_return_value(False)
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-        self.mock_os.set_os_path_exists_return_value(True)
-
-    def test_copy(self):
-        """
-        Test that deepcopy works correctly
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-            "--profile-models",
-            "vgg11",
-        ]
-        yaml_content = "model_repository: yaml_repository"
-        configA = self._evaluate_config(args, yaml_content)
-
-        configB = deepcopy(configA)
-
-        self._assert_equality_of_model_configs(
-            configA.get_all_config()["profile_models"],
-            configB.get_all_config()["profile_models"],
-        )
-
-        self.assertEqual(configA.run_config_search_mode, configB.run_config_search_mode)
-
-        configB.run_config_search_mode = "quick"
-
-        self.assertNotEqual(
-            configA.run_config_search_mode, configB.run_config_search_mode
-        )
-
-    def test_multi_model_search_mode(self):
-        """
-        Test that multi-model is only run in quick
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--profile-models",
-            "test_modelA,test_modelB",
-            "--run-config-profile-models-concurrently-enable",
-        ]
-
-        yaml_content = ""
-
-        # Tests the case where no search mode is specified (default is brute)
-        self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        # Brute should fail
-        new_args = list(args)
-        new_args.append("--run-config-search-mode")
-        new_args.append("brute")
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(new_args, yaml_content, subcommand="profile")
-
-        # Quick should pass
-        new_args = list(args)
-        new_args.append("--run-config-search-mode")
-        new_args.append("quick")
-
-        self._evaluate_config(new_args, yaml_content, subcommand="profile")
-
-    def test_quick_search_mode(self):
-        """
-        Test that only legal options are specified in quick search
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--profile-models",
-            "test_modelA",
-            "--run-config-search-mode",
-            "quick",
-        ]
-
-        yaml_content = ""
-
-        self._evaluate_config(args, yaml_content)
-
-        self._test_quick_search_with_rcs(
-            args,
-            yaml_content,
-            "--run-config-search-disable",
-            use_value=False,
-            use_list=False,
-        )
-        self._test_quick_search_with_rcs(
-            args, yaml_content, "--batch-sizes", use_value=False
-        )
-        self._test_quick_search_with_rcs(
-            args, yaml_content, "--concurrency", use_value=False
-        )
-
-    def test_quick_search_model_specific(self):
-        """
-        Test for illegal model specific options in the YAML during quick search
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--run-config-search-mode",
-            "quick",
-            "-f",
-            "path-to-config-file",
-        ]
-
-        yaml_content = """
-        profile_models:
-          model_1:
-            parameters:
-              concurrency: 1,2,4,128
-        """
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        yaml_content = """
-        profile_models:
-          model_1:
-            parameters:
-              batch size: 1,2,4,128
-        """
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        yaml_content = """
-        profile_models:
-          model_1:
-            model_config_parameters:
-              max_batch_size: 2
-        """
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        yaml_content = """
-        profile_models:
-          model_1:
-            model_config_parameters:
-              instance_group:
-                - kind: KIND_GPU
-        """
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        yaml_content = """
-        profile_models:
-          model_1:
-            model_config_parameters:
-               optimization:
-                 execution_accelerators:
-                   gpu_execution_accelerator:
-                     - name: tensorrt
-                       parameters:
-                       precision_mode: [FP16, FP32]
-        """
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-
-    def test_model_weightings_profile(self):
-        """
-        Test that weightings can be specified only per model in the YAML
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--run-config-search-mode",
-            "quick",
-            "-f",
-            "path-to-config-file",
-        ]
-
-        # Global weighting - illegal
-        yaml_content = """
-        weighting: 3
-
-        profile_models:
-          resnet50_libtorch:
-            objectives:
-              perf_latency_p99: 1
-          add_sub:
-            objectives:
-              perf_throughput: 1
-        """
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-
-        # Model specific weighting - legal
-        yaml_content = """
-        profile_models:
-          resnet50_libtorch:
-            weighting: 3
-            objectives:
-              perf_latency_p99: 1
-          add_sub:
-            weighting: 1
-            objectives:
-              perf_throughput: 1
-        """
-
-        self._evaluate_config(args, yaml_content, subcommand="profile")
-
-    def _test_quick_search_with_rcs(
-        self,
-        args: Namespace,
-        yaml_content: Optional[Dict[str, List]],
-        rcs_string: str,
-        use_value: bool = True,
-        use_list: bool = True,
-    ) -> None:
-        """
-        Tests that run-config-search options raise exceptions
-        in quick search mode
-        """
-        new_args = list(args)
-        new_args.append(rcs_string)
-        if use_value:
-            new_args.append("1")
-        elif use_list:
-            new_args.append(["1", "2", "4"])
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(new_args, yaml_content, subcommand="profile")
-
-    def test_bls_composing_models(self):
-        """
-        Test that BLS composing models can be specified
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--run-config-search-mode",
-            "quick",
-            "--profile-models",
-            "modelA",
-            "--bls-composing-models",
-            "modelA,modelB",
-        ]
-        yaml_content = ""
-
-        self._evaluate_config(args, yaml_content)
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--run-config-search-mode",
-            "quick",
-            "--profile-models",
-            "modelA",
-        ]
-
-        yaml_content = """
-        bls_composing_models: modelA,modelB
-        """
-
-        self._evaluate_config(args, yaml_content)
-
-    def test_bls_illegal_config_combinations(self):
-        """
-        Test BLS illegal config combinations
-        """
-        # BLS with brute search mode
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--run-config-search-mode",
-            "brute",
-            "--profile-models",
-            "modelA",
-            "--bls-composing-models",
-            "modelA,modelB",
-        ]
-        yaml_content = ""
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content)
-
-        # BLS with multiple models
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--run-config-search-mode",
-            "quick",
-            "--profile-models",
-            "modelA,modelB",
-            "--bls-composing-models",
-            "modelA,modelB",
-        ]
-        yaml_content = ""
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content)
-
-        # BLS with concurrent search mode
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--run-config-search-mode",
-            "quick",
-            "--profile-models",
-            "modelA",
-            "--bls-composing-models",
-            "modelA,modelB",
-            "--run-config-profile-models-concurrently-enable",
-        ]
-        yaml_content = ""
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content)
-
-    def test_concurrency_rate_request_config_combinations(self):
-        """
-        Test for concurrency with rate request conflicts
-        """
-        base_args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--profile-models",
-            "modelA",
-            "-c",
-            "1,2,3",
-        ]
-        yaml_content = ""
-
-        self._test_request_rate_config_conflicts(base_args, yaml_content)
-
-    def test_config_search_min_rate_request_config_combinations(self):
-        """
-        Test for concurrency min request with rate request conflicts
-        """
-        base_args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--profile-models",
-            "modelA",
-            "--run-config-search-min-concurrency",
-            "1",
-        ]
-        yaml_content = ""
-
-        self._test_request_rate_config_conflicts(base_args, yaml_content)
-
-    def test_config_search_max_rate_request_config_combinations(self):
-        """
-        Test for concurrency max request with rate request conflicts
-        """
-        base_args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "--profile-models",
-            "modelA",
-            "--run-config-search-max-concurrency",
-            "1",
-        ]
-        yaml_content = ""
-
-        self._test_request_rate_config_conflicts(base_args, yaml_content)
-
-    def test_mixing_request_rate_and_concurrency(self):
-        """
-        Test that mixing models parameters of request-rate and concurrency is illegal
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli-repository",
-            "-f",
-            "path-to-config-file",
-        ]
-
-        yaml_content = """
-        profile_models:
-
-          resnet50_libtorch:
-            parameters:
-              concurrency: 4
-          add_sub:
-            parameters:
-              request_rate: 16
-        """
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content, subcommand="profile")
-
-    def _test_request_rate_config_conflicts(
-        self, base_args: List[Any], yaml_content: str
-    ) -> None:
-        self._test_arg_conflict(
-            base_args, yaml_content, ["--request-rate-search-enable"]
-        )
-        self._test_arg_conflict(base_args, yaml_content, ["--request-rate", "1,2,3"])
-        self._test_arg_conflict(
-            base_args, yaml_content, ["--run-config-search-min-request-rate", "1"]
-        )
-        self._test_arg_conflict(
-            base_args, yaml_content, ["--run-config-search-max-request-rate", "1"]
-        )
-
-    def _test_arg_conflict(
-        self, base_args: List[Any], yaml_content: str, new_args: List[Any]
-    ) -> None:
-        args = base_args.copy() + new_args
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._evaluate_config(args, yaml_content)
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_constraint_manager.py b/src/c++/perf_analyzer/tests/test_constraint_manager.py
deleted file mode 100755
index f4e23e7df..000000000
--- a/src/c++/perf_analyzer/tests/test_constraint_manager.py
+++ /dev/null
@@ -1,620 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.constants import GLOBAL_CONSTRAINTS_KEY
-from model_analyzer.result.constraint_manager import ConstraintManager
-from model_analyzer.result.model_constraints import ModelConstraints
-
-from .common import test_result_collector as trc
-from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
-
-
-class TestConstraintManager(trc.TestResultCollector):
-    def tearDown(self):
-        patch.stopall()
-
-    def test_single_model_no_constraints(self):
-        """
-        Test that constraints are empty
-        """
-        config = self._create_single_model_no_constraints()
-        constraint_manager = ConstraintManager(config)
-        constraints = constraint_manager.get_constraints_for_all_models()
-
-        self.assertEqual(constraints["model_A"], ModelConstraints({}))
-        self.assertEqual(constraints[GLOBAL_CONSTRAINTS_KEY], ModelConstraints({}))
-
-    def test_single_model_with_constraints(self):
-        """
-        Test that model specific constraints are set
-        """
-        config = self._create_single_model_with_constraints()
-        constraint_manager = ConstraintManager(config)
-        constraints = constraint_manager.get_constraints_for_all_models()
-
-        self.assertEqual(
-            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 100}})
-        )
-        self.assertEqual(constraints[GLOBAL_CONSTRAINTS_KEY], ModelConstraints({}))
-
-    def test_single_model_with_global_constraints(self):
-        """
-        Test that global constraints are attributed to a model
-        """
-        config = self._create_single_model_global_constraints()
-        constraint_manager = ConstraintManager(config)
-        constraints = constraint_manager.get_constraints_for_all_models()
-
-        self.assertEqual(
-            constraints["model_A"], ModelConstraints({"perf_throughput": {"min": 100}})
-        )
-        self.assertEqual(
-            constraints[GLOBAL_CONSTRAINTS_KEY],
-            ModelConstraints({"perf_throughput": {"min": 100}}),
-        )
-
-    def test_single_model_with_both_constraints(self):
-        """
-        Test that model specific constraints override global
-        """
-        config = self._create_single_model_both_constraints()
-        constraint_manager = ConstraintManager(config)
-        constraints = constraint_manager.get_constraints_for_all_models()
-
-        self.assertEqual(
-            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 50}})
-        )
-        self.assertEqual(
-            constraints[GLOBAL_CONSTRAINTS_KEY],
-            ModelConstraints({"perf_latency_p99": {"max": 100}}),
-        )
-
-    def test_multi_model_with_no_global_constraints(self):
-        """
-        Test multi-model with only model constraints and no global constraints
-        """
-        config = self._create_multi_model_with_no_global_constraints()
-        constraint_manager = ConstraintManager(config)
-        constraints = constraint_manager.get_constraints_for_all_models()
-
-        self.assertEqual(
-            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 50}})
-        )
-        self.assertEqual(
-            constraints["model_B"], ModelConstraints({"perf_throughput": {"min": 100}})
-        )
-
-    def test_multi_model_with_matching_global_constraints(self):
-        """
-        Test multi-model with Global constraints and individual overrides
-        """
-        config = self._create_multi_model_with_matching_global_constraints()
-        constraint_manager = ConstraintManager(config)
-        constraints = constraint_manager.get_constraints_for_all_models()
-
-        self.assertEqual(
-            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 75}})
-        )
-        self.assertEqual(
-            constraints["model_B"], ModelConstraints({"perf_throughput": {"min": 125}})
-        )
-        self.assertEqual(
-            constraints[GLOBAL_CONSTRAINTS_KEY],
-            ModelConstraints(
-                {"perf_throughput": {"min": 100}, "gpu_used_memory": {"max": 1000}}
-            ),
-        )
-
-    def test_multi_model_with_different_global_constrants(self):
-        """
-        Test multi-model with different global and individual model constraints
-        """
-        config = self._create_multi_model_with_different_global_constrants()
-        constraint_manager = ConstraintManager(config)
-        constraints = constraint_manager.get_constraints_for_all_models()
-
-        self.assertEqual(
-            constraints["model_A"], ModelConstraints({"perf_latency_p99": {"max": 100}})
-        )
-        self.assertEqual(
-            constraints["model_B"], ModelConstraints({"perf_throughput": {"min": 150}})
-        )
-        self.assertEqual(
-            constraints[GLOBAL_CONSTRAINTS_KEY],
-            ModelConstraints({"gpu_used_memory": {"max": 2000}}),
-        )
-
-    def test_single_model_max_constraint_checks(self):
-        """
-        Test that satisfies_constraints works for a single model
-        with a max style constraint
-        """
-        config = self._create_single_model_with_constraints()
-        constraint_manager = ConstraintManager(config)
-
-        # Constraint is P99 Latency max of 100
-        rcm = self._construct_rcm({"perf_latency_p99": 101}, constraint_manager)
-        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
-
-        rcm = self._construct_rcm({"perf_latency_p99": 100}, constraint_manager)
-        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
-
-        rcm = self._construct_rcm({"perf_latency_p99": 99}, constraint_manager)
-        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
-
-    def test_single_model_min_constraint_checks(self):
-        """
-        Test that satisfies_constraints works for a single model
-        with a min style constraint
-        """
-        config = self._create_single_model_global_constraints()
-        constraint_manager = ConstraintManager(config)
-
-        # Constraint is throughput min of 100
-        rcm = self._construct_rcm({"perf_throughput": 101}, constraint_manager)
-        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
-
-        rcm = self._construct_rcm({"perf_throughput": 100}, constraint_manager)
-        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
-
-        rcm = self._construct_rcm({"perf_throughput": 99}, constraint_manager)
-        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
-
-    def test_multi_model_max_constraint_checks(self):
-        """
-        Test that satisfies_constraints works for multi-model
-        with a max style constraints
-        """
-        # Constraints are:
-        #  Model A: P99 Latency max of 50
-        #  Model B: Throughput min of 100
-        constraint_manager = ConstraintManager(
-            config=self._create_multi_model_with_no_global_constraints()
-        )
-
-        # Model A & B are both at boundaries
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 50, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 100},
-            ],
-            constraint_manager,
-        )
-        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
-
-        # Model A exceeds latency
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 51, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 100},
-            ],
-            constraint_manager,
-        )
-        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
-
-        # Constraints are:
-        #  Model A: P99 Latency max of 75
-        #  Model B: Throughput min of 125
-        constraint_manager = ConstraintManager(
-            config=self._create_multi_model_with_matching_global_constraints()
-        )
-
-        # Model A & B are both at boundaries
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 75, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 125},
-            ],
-            constraint_manager,
-        )
-        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
-
-        # Model A exceeds latency
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 78, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 125},
-            ],
-            constraint_manager,
-        )
-        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
-
-        # Constraints are:
-        #  Model A: P99 Latency max of 100
-        #  Model B: Throughput min of 150
-        constraint_manager = ConstraintManager(
-            config=self._create_multi_model_with_different_global_constrants()
-        )
-
-        # Model A & B are both at boundaries
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 100, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 150},
-            ],
-            constraint_manager,
-        )
-        self.assertTrue(constraint_manager.satisfies_constraints(rcm))
-
-        # Model A exceeds latency
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 105, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 150},
-            ],
-            constraint_manager,
-        )
-        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
-
-    def test_multi_model_min_constraint_checks(self):
-        """
-        Test that satisfies_constraints works for multi-model
-        with a min style constraints
-        """
-        # Constraints are:
-        #  Model A: P99 Latency max of 50
-        #  Model B: Throughput min of 100
-        constraint_manager = ConstraintManager(
-            config=self._create_multi_model_with_no_global_constraints()
-        )
-
-        # Model B doesn't have enough throughput
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 50, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 99},
-            ],
-            constraint_manager,
-        )
-        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
-
-        # Constraints are:
-        #  Model A: P99 Latency max of 75
-        #  Model B: Throughput min of 125
-        constraint_manager = ConstraintManager(
-            config=self._create_multi_model_with_matching_global_constraints()
-        )
-
-        # Model B doesn't have enough throughput
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 75, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 120},
-            ],
-            constraint_manager,
-        )
-        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
-
-        # Constraints are:
-        #  Model A: P99 Latency max of 100
-        #  Model B: Throughput min of 150
-        constraint_manager = ConstraintManager(
-            config=self._create_multi_model_with_different_global_constrants()
-        )
-
-        # Model B doesn't have enough throughput
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 100, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 140},
-            ],
-            constraint_manager,
-        )
-        self.assertFalse(constraint_manager.satisfies_constraints(rcm))
-
-    def test_single_model_max_failure_percentage(self):
-        """
-        Test that constraint_failure_percentage works for a single model
-        with a max style constraint
-        """
-        config = self._create_single_model_with_constraints()
-        constraint_manager = ConstraintManager(config)
-
-        # Constraint is P99 Latency max of 100
-        rcm = self._construct_rcm({"perf_latency_p99": 225}, constraint_manager)
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 125)
-        rcm = self._construct_rcm({"perf_latency_p99": 150}, constraint_manager)
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
-
-        rcm = self._construct_rcm({"perf_latency_p99": 100}, constraint_manager)
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
-
-        rcm = self._construct_rcm({"perf_latency_p99": 99}, constraint_manager)
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
-
-    def test_single_model_min_failure_percentage(self):
-        """
-        Test that constraint_failure_percentage works for a single model
-        with a min style constraint
-        """
-
-        config = self._create_single_model_global_constraints()
-        constraint_manager = ConstraintManager(config)
-
-        # Constraint is throughput min of 100
-        rcm = self._construct_rcm({"perf_throughput": 25}, constraint_manager)
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 75)
-        rcm = self._construct_rcm({"perf_throughput": 50}, constraint_manager)
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
-
-        rcm = self._construct_rcm({"perf_throughput": 100}, constraint_manager)
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
-
-        rcm = self._construct_rcm({"perf_throughput": 101}, constraint_manager)
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
-
-    def test_multi_model_failure_percentage(self):
-        """
-        Test that constraint_failure_percentage works for multi-model
-        with a max style constraints
-        """
-        # Constraints are:
-        #  Model A: P99 Latency max of 50
-        #  Model B: Throughput min of 100
-        constraint_manager = ConstraintManager(
-            config=self._create_multi_model_with_no_global_constraints()
-        )
-
-        # Model A & B are both at boundaries
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 50, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 100},
-            ],
-            constraint_manager,
-        )
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
-
-        # Model A exceeds latency, Model B misses on throughput - each by 20%
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 60, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 80},
-            ],
-            constraint_manager,
-        )
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 40)
-
-        # Model A exceeds latency by 40%, Model B misses on throughput by 10%
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 70, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 90},
-            ],
-            constraint_manager,
-        )
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
-
-        # Constraints are:
-        #  Model A: P99 Latency max of 75
-        #  Model B: Throughput min of 125
-        constraint_manager = ConstraintManager(
-            config=self._create_multi_model_with_matching_global_constraints()
-        )
-
-        # Model A & B are both at boundaries
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 75, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 125},
-            ],
-            constraint_manager,
-        )
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
-
-        # Model A exceeds latency, Model B misses on throughput - each by 25%
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 93.75, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 93.75},
-            ],
-            constraint_manager,
-        )
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
-
-        # Model A exceeds latency by 30%, Model B misses on throughput by 20%
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 97.5, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 100},
-            ],
-            constraint_manager,
-        )
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 50)
-
-        # Constraints are:
-        #  Model A: P99 Latency max of 100
-        #  Model B: Throughput min of 150
-        constraint_manager = ConstraintManager(
-            config=self._create_multi_model_with_different_global_constrants()
-        )
-
-        # Model A & B are both at boundaries
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 100, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 150},
-            ],
-            constraint_manager,
-        )
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 0)
-
-        # Model A exceeds latency, Model B misses on throughput - each by 30%
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 130, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 105},
-            ],
-            constraint_manager,
-        )
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 60)
-
-        # Model A exceeds latency by 35%, Model B misses on throughput by 25%
-        rcm = self._construct_mm_rcm(
-            [
-                {"perf_latency_p99": 135, "perf_throughput": 0},
-                {"perf_latency_p99": 0, "perf_throughput": 112.5},
-            ],
-            constraint_manager,
-        )
-        self.assertEqual(constraint_manager.constraint_failure_percentage(rcm), 60)
-
-    def _create_single_model_no_constraints(self):
-        args = self._create_args()
-        yaml_str = """
-            profile_models:
-              model_A
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        return config
-
-    def _create_single_model_with_constraints(self):
-        args = self._create_args()
-        yaml_str = """
-            profile_models:
-              model_A:
-                constraints:
-                  perf_latency_p99:
-                    max: 100
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        return config
-
-    def _create_single_model_global_constraints(self):
-        args = self._create_args()
-        yaml_str = """
-            profile_models:
-              model_A
-
-            constraints:
-                perf_throughput:
-                  min: 100
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        return config
-
-    def _create_single_model_both_constraints(self):
-        args = self._create_args()
-        yaml_str = """
-            profile_models:
-              model_A:
-                constraints:
-                  perf_latency_p99:
-                    max: 50
-
-            constraints:
-                perf_latency_p99:
-                  max: 100
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        return config
-
-    def _create_multi_model_with_no_global_constraints(self):
-        args = self._create_args()
-        yaml_str = """
-            profile_models:
-              model_A:
-                constraints:
-                  perf_latency_p99:
-                    max: 50
-              model_B:
-                constraints:
-                  perf_throughput:
-                    min: 100
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        return config
-
-    def _create_multi_model_with_matching_global_constraints(self):
-        args = self._create_args()
-        yaml_str = """
-            profile_models:
-              model_A:
-                constraints:
-                  perf_latency_p99:
-                    max: 75
-              model_B:
-                constraints:
-                  perf_throughput:
-                    min: 125
-            constraints:
-                perf_throughput:
-                  min: 100
-                gpu_used_memory:
-                  max: 1000
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        return config
-
-    def _create_multi_model_with_different_global_constrants(self):
-        args = self._create_args()
-        yaml_str = """
-            profile_models:
-              model_A:
-                constraints:
-                  perf_latency_p99:
-                    max: 100
-              model_B:
-                constraints:
-                  perf_throughput:
-                    min: 150
-            constraints:
-                gpu_used_memory:
-                  max: 2000
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        return config
-
-    def _construct_rcm(self, non_gpu_metric_values, constraint_manager):
-        rcm = construct_run_config_measurement(
-            model_name=MagicMock(),
-            model_config_names=["model_A_config_name_0"],
-            constraint_manager=constraint_manager,
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values=MagicMock(),
-            non_gpu_metric_values=[non_gpu_metric_values],
-        )
-
-        return rcm
-
-    def _construct_mm_rcm(self, non_gpu_metric_values, constraint_manager):
-        rcm = construct_run_config_measurement(
-            model_name=MagicMock(),
-            model_config_names=["model_A_config_name_A", "model_B_config_name_B"],
-            constraint_manager=constraint_manager,
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values=MagicMock(),
-            non_gpu_metric_values=non_gpu_metric_values,
-        )
-
-        return rcm
-
-    def _create_args(self):
-        return ["model-analyzer", "profile", "-f", "config.yml", "-m", "."]
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_coordinate.py b/src/c++/perf_analyzer/tests/test_coordinate.py
deleted file mode 100755
index 156c8bb2f..000000000
--- a/src/c++/perf_analyzer/tests/test_coordinate.py
+++ /dev/null
@@ -1,138 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from copy import deepcopy
-
-from model_analyzer.config.generate.coordinate import Coordinate
-
-from .common import test_result_collector as trc
-
-
-class TestCoordinate(trc.TestResultCollector):
-    def test_basic(self):
-        x = Coordinate([0, 0])
-
-        self.assertEqual(x[0], 0)
-        self.assertEqual(x[1], 0)
-
-        x[0] = 2
-        x[1] = 4
-
-        self.assertEqual(x[0], 2)
-        self.assertEqual(x[1], 4)
-
-    def test_out_of_range(self):
-        x = Coordinate([0, 0])
-        with self.assertRaises(IndexError):
-            _ = x[2]
-
-    def test_assignment(self):
-        c1 = Coordinate([2, 4])
-
-        c2 = deepcopy(c1)
-
-        self.assertEqual(c2[0], 2)
-        self.assertEqual(c2[1], 4)
-
-        # Confirm c1 unchanged by c2 update
-        c2[0] = 5
-
-        self.assertEqual(c1[0], 2)
-        self.assertEqual(c1[1], 4)
-
-    def test_coordinate_addition(self):
-        c1 = Coordinate([2, 4])
-        c2 = Coordinate([5, 1])
-
-        c3 = c1 + c2
-        self.assertEqual(c3[0], 7)
-        self.assertEqual(c3[1], 5)
-
-    def test_integer_addition(self):
-        c1 = Coordinate([2, 4])
-        c2 = c1 + 3
-
-        self.assertEqual(c2[0], 5)
-        self.assertEqual(c2[1], 7)
-
-    def test_float_addition(self):
-        c1 = Coordinate([2, 4])
-
-        c2 = c1 + 3.5
-
-        self.assertEqual(c2[0], 5.5)
-        self.assertEqual(c2[1], 7.5)
-
-    def test_division(self):
-        c1 = Coordinate([7, 6])
-
-        c1 /= 2
-
-        self.assertEqual(c1[0], 3.5)
-        self.assertEqual(c1[1], 3)
-
-    def test_multiplication(self):
-        c1 = Coordinate([7, 6])
-
-        c1 *= 2
-
-        self.assertEqual(c1[0], 14)
-        self.assertEqual(c1[1], 12)
-
-    def test_round(self):
-        c1 = Coordinate([0.1, 4.6, 3.9])
-        c1.round()
-        self.assertEqual(c1, Coordinate([0, 5, 4]))
-
-    def test_iterator(self):
-        c1 = Coordinate([2, 4, 1, 7])
-
-        values = []
-        for x in c1:
-            values.append(x)
-
-        self.assertEqual(values, [2, 4, 1, 7])
-
-    def test_enumerate(self):
-        c1 = Coordinate([2, 4, 1, 7])
-
-        indexes = []
-        values = []
-        for i, x in enumerate(c1):
-            indexes.append(i)
-            values.append(x)
-
-        self.assertEqual(indexes, [0, 1, 2, 3])
-        self.assertEqual(values, [2, 4, 1, 7])
-
-    def test_stringification(self):
-        c1 = Coordinate([2, 4, 1, 7])
-        self.assertEqual("[2, 4, 1, 7]", c1.__str__())
-
-    def test_compare(self):
-        c1 = Coordinate([1, 3, 2])
-        c2 = Coordinate([1, 3, 2])
-        c3 = Coordinate([1, 3, 1])
-        c4 = Coordinate([1, 3, 3])
-        c5 = Coordinate([2, 1, 3])
-        c6 = Coordinate([0, 1, 3])
-
-        self.assertEqual(c1, c2)
-        self.assertGreater(c1, c3)
-        self.assertLess(c3, c1)
-        self.assertLess(c1, c4)
-        self.assertLess(c1, c5)
-        self.assertGreater(c1, c6)
diff --git a/src/c++/perf_analyzer/tests/test_coordinate_data.py b/src/c++/perf_analyzer/tests/test_coordinate_data.py
deleted file mode 100755
index a668fd3d1..000000000
--- a/src/c++/perf_analyzer/tests/test_coordinate_data.py
+++ /dev/null
@@ -1,123 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from unittest.mock import MagicMock
-
-from model_analyzer.config.generate.coordinate import Coordinate
-from model_analyzer.config.generate.coordinate_data import CoordinateData
-
-from .common import test_result_collector as trc
-from .common.test_utils import construct_run_config_measurement
-
-
-class TestCoordinateData(trc.TestResultCollector):
-    def _construct_rcm(
-        self, throughput: float, latency: float, config_name: str, model_name: str
-    ):
-        model_config_name = [config_name]
-
-        # yapf: disable
-        non_gpu_metric_values = [{
-            "perf_throughput": throughput,
-            "perf_latency_avg": latency
-        }]
-        # yapf: enable
-
-        metric_objectives = [{"perf_throughput": 1}]
-        weights = [1]
-
-        rcm = construct_run_config_measurement(
-            model_name=model_name,
-            model_config_names=model_config_name,
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values={},
-            non_gpu_metric_values=non_gpu_metric_values,
-            metric_objectives=metric_objectives,
-            model_config_weights=weights,
-        )
-        return rcm
-
-    def test_basic(self):
-        result_data = CoordinateData()
-
-        coordinate = Coordinate([0, 0, 0])
-        self.assertEqual(result_data.get_measurement(coordinate), None)
-        self.assertEqual(result_data.get_visit_count(coordinate), 0)
-        self.assertEqual(result_data.is_measured(coordinate), False)
-        self.assertEqual(result_data.has_valid_measurement(coordinate), False)
-
-    def test_visit_count(self):
-        result_data = CoordinateData()
-
-        coordinate1 = Coordinate([0, 0, 0])
-        coordinate2 = Coordinate([0, 4, 1])
-
-        result_data.increment_visit_count(coordinate1)
-        self.assertEqual(1, result_data.get_visit_count(coordinate1))
-
-        result_data.increment_visit_count(coordinate2)
-        self.assertEqual(1, result_data.get_visit_count(coordinate2))
-
-        result_data.increment_visit_count(coordinate1)
-        result_data.increment_visit_count(coordinate1)
-        self.assertEqual(3, result_data.get_visit_count(coordinate1))
-        self.assertEqual(1, result_data.get_visit_count(coordinate2))
-
-    def test_measurement(self):
-        """
-        Test if CoordinateData can properly set and get measurements
-
-        Also confirm that is_measured() and has_valid_measurement() work properly
-        """
-        coordinate_data = CoordinateData()
-
-        coordinate0 = Coordinate([0, 0, 0])
-        coordinate1 = Coordinate([0, 4, 1])
-        coordinate2 = Coordinate([1, 2, 3])
-
-        rcm0 = self._construct_rcm(
-            10, 5, config_name="modelA_config_0", model_name="modelA"
-        )
-        rcm1 = self._construct_rcm(
-            20, 8, config_name="modelB_config_0", model_name="modelB"
-        )
-        rcm2 = None
-
-        coordinate_data.set_measurement(coordinate0, rcm0)
-        coordinate_data.set_measurement(coordinate1, rcm1)
-        coordinate_data.set_measurement(coordinate2, rcm2)
-
-        self.assertEqual(coordinate_data.is_measured(coordinate0), True)
-        self.assertEqual(coordinate_data.is_measured(coordinate1), True)
-        self.assertEqual(coordinate_data.is_measured(coordinate2), True)
-
-        self.assertEqual(coordinate_data.has_valid_measurement(coordinate0), True)
-        self.assertEqual(coordinate_data.has_valid_measurement(coordinate1), True)
-        self.assertEqual(coordinate_data.has_valid_measurement(coordinate2), False)
-
-        measurement0 = coordinate_data.get_measurement(coordinate0)
-        self.assertEqual("modelA_config_0", measurement0.model_variants_name())
-        self.assertEqual(10, measurement0.get_non_gpu_metric_value("perf_throughput"))
-        self.assertEqual(5, measurement0.get_non_gpu_metric_value("perf_latency_avg"))
-        self.assertTrue(coordinate_data.is_measured(coordinate0))
-
-        measurement1 = coordinate_data.get_measurement(coordinate1)
-        self.assertEqual("modelB_config_0", measurement1.model_variants_name())
-        self.assertEqual(20, measurement1.get_non_gpu_metric_value("perf_throughput"))
-        self.assertEqual(8, measurement1.get_non_gpu_metric_value("perf_latency_avg"))
-
-        measurement2 = coordinate_data.get_measurement(coordinate2)
-        self.assertEqual(measurement2, None)
diff --git a/src/c++/perf_analyzer/tests/test_cpu_monitor.py b/src/c++/perf_analyzer/tests/test_cpu_monitor.py
deleted file mode 100755
index 030fa9816..000000000
--- a/src/c++/perf_analyzer/tests/test_cpu_monitor.py
+++ /dev/null
@@ -1,111 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import time
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.device.gpu_device import GPUDevice
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.monitor.cpu_monitor import CPUMonitor
-from model_analyzer.record.types.cpu_available_ram import CPUAvailableRAM
-from model_analyzer.record.types.cpu_used_ram import CPUUsedRAM
-from model_analyzer.triton.server.server_config import TritonServerConfig
-from model_analyzer.triton.server.server_factory import TritonServerFactory
-
-from .common import test_result_collector as trc
-from .mocks.mock_server_local import MockServerLocalMethods
-
-MODEL_REPOSITORY_PATH = "test_repo"
-TRITON_LOCAL_BIN_PATH = "test_bin_path/tritonserver"
-
-
-class TestCPUMonitor(trc.TestResultCollector):
-    def setUp(self):
-        self.server_local_mock = MockServerLocalMethods()
-        self.server_local_mock.start()
-
-    def test_record_cpu_memory(self):
-        server_config = TritonServerConfig()
-        server_config["model-repository"] = MODEL_REPOSITORY_PATH
-        gpus = [GPUDevice("TEST_DEVICE_NAME", 0, "TEST_PCI_BUS_ID", "TEST_UUID")]
-
-        frequency = 1
-        monitoring_time = 0.1
-        metrics = [CPUAvailableRAM, CPUUsedRAM]
-
-        server = TritonServerFactory.create_server_local(
-            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
-        )
-
-        # Start triton and monitor
-        server.start()
-        cpu_monitor = CPUMonitor(server, frequency, metrics)
-        cpu_monitor.start_recording_metrics()
-        time.sleep(monitoring_time)
-        records = cpu_monitor.stop_recording_metrics()
-
-        # Assert library calls
-        self.server_local_mock.assert_cpu_stats_called()
-
-        # Assert instance types
-        for record in records:
-            self.assertIsInstance(record.value(), float)
-            self.assertIsInstance(record.timestamp(), int)
-
-        # The number of records should be dividable by number of metrics
-        self.assertEqual(len(records) % len(metrics), 0)
-        self.assertGreater(len(records), 0)
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            cpu_monitor.stop_recording_metrics()
-
-        cpu_monitor.destroy()
-        server.stop()
-
-    def test_monitor_disable(self):
-        server_config = TritonServerConfig()
-        server_config["model-repository"] = MODEL_REPOSITORY_PATH
-        gpus = [GPUDevice("TEST_DEVICE_NAME", 0, "TEST_PCI_BUS_ID", "TEST_UUID")]
-
-        frequency = 1
-        monitoring_time = 0.1
-        metrics = []
-
-        server = TritonServerFactory.create_server_local(
-            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
-        )
-
-        # Start triton and monitor
-        server.start()
-        cpu_monitor = CPUMonitor(server, frequency, metrics)
-        cpu_monitor.start_recording_metrics()
-        time.sleep(monitoring_time)
-        cpu_monitor.stop_recording_metrics()
-
-        # Assert no library calls
-        self.server_local_mock.assert_cpu_stats_not_called()
-
-        cpu_monitor.destroy()
-        server.stop()
-
-    def tearDown(self):
-        patch.stopall()
-        self.server_local_mock.stop()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_dcgm_monitor.py b/src/c++/perf_analyzer/tests/test_dcgm_monitor.py
deleted file mode 100755
index ebdbee037..000000000
--- a/src/c++/perf_analyzer/tests/test_dcgm_monitor.py
+++ /dev/null
@@ -1,155 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import time
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.device.gpu_device import GPUDevice
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.monitor.dcgm.dcgm_monitor import DCGMMonitor
-from model_analyzer.record.types.gpu_free_memory import GPUFreeMemory
-from model_analyzer.record.types.gpu_power_usage import GPUPowerUsage
-from model_analyzer.record.types.gpu_used_memory import GPUUsedMemory
-from model_analyzer.record.types.gpu_utilization import GPUUtilization
-
-from .common import test_result_collector as trc
-from .mocks.mock_dcgm import MockDCGM
-from .mocks.mock_dcgm_agent import TEST_PCI_BUS_ID, TEST_UUID
-from .mocks.mock_dcgm_field_group_watcher import TEST_RECORD_VALUE
-from .mocks.mock_numba import MockNumba
-
-TEST_DEVICE_NAME = "TEST_DEVICE_NAME"
-TEST_DEVICE_ID = 0
-
-
-class TestDCGMMonitor(trc.TestResultCollector):
-    def setUp(self):
-        self.mock_dcgm = MockDCGM()
-        self.mock_numba = MockNumba(
-            mock_paths=[
-                "model_analyzer.device.gpu_device_factory",
-            ]
-        )
-        self.mock_dcgm.start()
-        self.mock_numba.start()
-
-        self._gpus = [
-            GPUDevice(TEST_DEVICE_NAME, TEST_DEVICE_ID, TEST_PCI_BUS_ID, TEST_UUID)
-        ]
-
-    def test_record_memory(self):
-        # One measurement every 0.01 seconds
-        frequency = 1
-        monitoring_time = 0.1
-        metrics = [GPUUsedMemory, GPUFreeMemory]
-        dcgm_monitor = DCGMMonitor(self._gpus, frequency, metrics)
-        dcgm_monitor.start_recording_metrics()
-        time.sleep(monitoring_time)
-        records = dcgm_monitor.stop_recording_metrics()
-
-        # Assert instance types
-        for record in records:
-            self.assertIsInstance(record.device_uuid(), str)
-            self.assertIsInstance(record.value(), float)
-            self.assertEqual(record.value(), TEST_RECORD_VALUE)
-            self.assertIsInstance(record.timestamp(), int)
-
-        # The number of records should be dividable by number of metrics
-        self.assertEqual(len(records) % len(metrics), 0)
-        self.assertGreater(len(records), 0)
-        self.assertGreaterEqual(
-            records[-1].timestamp() - records[0].timestamp(), monitoring_time
-        )
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            dcgm_monitor.stop_recording_metrics()
-
-        dcgm_monitor.destroy()
-
-        metrics = ["UndefinedTag"]
-        with self.assertRaises(TritonModelAnalyzerException):
-            DCGMMonitor(self._gpus, frequency, metrics)
-
-    def test_record_power(self):
-        # One measurement every 0.01 seconds
-        frequency = 1
-        monitoring_time = 1.1
-        metrics = [GPUPowerUsage]
-        dcgm_monitor = DCGMMonitor(self._gpus, frequency, metrics)
-        dcgm_monitor.start_recording_metrics()
-        time.sleep(monitoring_time)
-        records = dcgm_monitor.stop_recording_metrics()
-
-        # Assert instance types
-        for record in records:
-            self.assertIsInstance(record.device_uuid(), str)
-            self.assertIsInstance(record.value(), float)
-            self.assertEqual(record.value(), TEST_RECORD_VALUE)
-            self.assertIsInstance(record.timestamp(), int)
-
-        # The number of records should be dividable by number of metrics
-        self.assertEqual(len(records) % len(metrics), 0)
-        self.assertGreater(len(records), 0)
-        self.assertGreaterEqual(
-            records[-1].timestamp() - records[0].timestamp(), monitoring_time
-        )
-
-        dcgm_monitor.destroy()
-
-    def test_record_utilization(self):
-        # One measurement every 0.01 seconds
-        frequency = 1
-        monitoring_time = 1.1
-        metrics = [GPUUtilization]
-        dcgm_monitor = DCGMMonitor(self._gpus, frequency, metrics)
-        dcgm_monitor.start_recording_metrics()
-        time.sleep(monitoring_time)
-        records = dcgm_monitor.stop_recording_metrics()
-
-        # Assert instance types
-        for record in records:
-            self.assertIsInstance(record.device_uuid(), str)
-            self.assertIsInstance(record.value(), float)
-            self.assertLessEqual(record.value(), 100)
-            self.assertEqual(record.value(), TEST_RECORD_VALUE)
-            self.assertIsInstance(record.timestamp(), int)
-
-        # The number of records should be dividable by number of metrics
-        self.assertEqual(len(records) % len(metrics), 0)
-        self.assertGreater(len(records), 0)
-        self.assertGreaterEqual(
-            records[-1].timestamp() - records[0].timestamp(), monitoring_time
-        )
-
-        dcgm_monitor.destroy()
-
-    def test_immediate_start_stop(self):
-        frequency = 1
-        metrics = [GPUUsedMemory, GPUFreeMemory]
-        dcgm_monitor = DCGMMonitor(self._gpus, frequency, metrics)
-        dcgm_monitor.start_recording_metrics()
-        dcgm_monitor.stop_recording_metrics()
-        dcgm_monitor.destroy()
-
-    def tearDown(self):
-        patch.stopall()
-        self.mock_dcgm.stop()
-        self.mock_numba.stop()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_ensemble_report_manager.py b/src/c++/perf_analyzer/tests/test_ensemble_report_manager.py
deleted file mode 100755
index a4f04cc89..000000000
--- a/src/c++/perf_analyzer/tests/test_ensemble_report_manager.py
+++ /dev/null
@@ -1,178 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import shutil
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.reports.report_manager import ReportManager
-from model_analyzer.result.result_manager import ResultManager
-from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
-
-from .common import test_result_collector as trc
-from .common.test_utils import ROOT_DIR, evaluate_mock_config
-
-
-class TestEnsembleReportManager(trc.TestResultCollector):
-    def _init_managers(
-        self,
-        models="test_model",
-        num_configs_per_model=10,
-        mode="online",
-        subcommand="profile",
-    ):
-        args = ["model-analyzer", subcommand, "-f", "path-to-config-file"]
-        args.extend(["--checkpoint-directory", f"{ROOT_DIR}/ensemble-ckpt"])
-
-        if subcommand == "profile":
-            args.extend(["--profile-models", models])
-            args.extend(["--model-repository", "/tmp"])
-        else:
-            args.extend(["--report-model-configs", models])
-
-        yaml_str = f"""
-            client_protocol: grpc
-            export_path: {ROOT_DIR}
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand=subcommand)
-        state_manager = AnalyzerStateManager(config=config, server=None)
-
-        state_manager.load_checkpoint(checkpoint_required=True)
-        gpu_info = {"gpu_uuid": {"name": "fake_gpu_name", "total_memory": 1024000000}}
-        self.result_manager = ResultManager(
-            config=config, state_manager=state_manager, constraint_manager=MagicMock()
-        )
-        self.report_manager = ReportManager(
-            mode=mode,
-            config=config,
-            gpu_info=gpu_info,
-            result_manager=self.result_manager,
-            constraint_manager=MagicMock(),
-        )
-
-    def test_ensemble_summary(self):
-        """
-        Ensures the summary report sentence and table are accurate for a basic ensemble model (loaded from a checkpoint)
-        """
-        self._init_managers(models="ensemble_add_sub", subcommand="profile")
-
-        self.report_manager.create_summaries()
-
-        expected_summary_sentence = (
-            "In 21 measurements across 5 configurations, <strong>ensemble_add_sub_config_1</strong> "
-            "is <strong>1%</strong> better than the default configuration at maximizing throughput, "
-            "under the given constraints, on GPU(s) TITAN RTX.<BR><BR><strong>ensemble_add_sub_config_1</strong> is comprised of the following composing models: "
-            "<UL> <LI> <strong>add_config_0</strong>: "
-            "1 GPU instance with a max batch size of 0 on platform python </LI><LI> <strong>sub_config_1</strong>: "
-            "2 GPU instances with a max batch size of 0 on platform python </LI> </UL>"
-        )
-
-        summary_table, summary_sentence = self.report_manager._build_summary_table(
-            report_key="ensemble_add_sub",
-            num_measurements=21,
-            num_configurations=5,
-            gpu_name="TITAN RTX",
-            report_gpu_metrics=True,
-        )
-
-        self.assertEqual(summary_sentence, expected_summary_sentence)
-
-        # Check the first row of the table
-        self.assertEqual(
-            summary_table._rows[0][0], "ensemble_add_sub_config_1"
-        )  # model config name
-        self.assertEqual(summary_table._rows[0][1], "(0, 0)")  # max batch size
-        self.assertEqual(
-            summary_table._rows[0][2], "(Disabled, Disabled)"
-        )  # dynamic batching
-        self.assertEqual(summary_table._rows[0][3], "(1:GPU, 2:GPU)")  # instance count
-        self.assertEqual(summary_table._rows[0][4], "3.655")  # p99 latency
-        self.assertEqual(summary_table._rows[0][5], "2865.88")  # throughput
-        self.assertEqual(summary_table._rows[0][6], 870)  # max gpu memory
-        self.assertEqual(summary_table._rows[0][7], 0.0)  # GPU utilization
-
-    def test_ensemble_summary_cpu_only(self):
-        """
-        Ensures the summary report sentence and table are accurate for a basic ensemble model (loaded from a checkpoint)
-        when the cpu only flag is set
-        """
-
-        self._init_managers(models="ensemble_add_sub", subcommand="profile")
-
-        self.report_manager.create_summaries()
-
-        expected_summary_sentence = (
-            "In 21 measurements across 5 configurations, <strong>ensemble_add_sub_config_1</strong> "
-            "is <strong>1%</strong> better than the default configuration at maximizing throughput, "
-            "under the given constraints.<BR><BR><strong>ensemble_add_sub_config_1</strong> is comprised of the following composing models: "
-            "<UL> <LI> <strong>add_config_0</strong>: "
-            "1 GPU instance with a max batch size of 0 on platform python </LI><LI> <strong>sub_config_1</strong>: "
-            "2 GPU instances with a max batch size of 0 on platform python </LI> </UL>"
-        )
-
-        summary_table, summary_sentence = self.report_manager._build_summary_table(
-            report_key="ensemble_add_sub",
-            num_measurements=21,
-            num_configurations=5,
-            gpu_name="TITAN RTX",
-            report_gpu_metrics=False,
-        )
-
-        self.assertEqual(summary_sentence, expected_summary_sentence)
-
-        # Check the first row of the table
-        self.assertEqual(
-            summary_table._rows[0][0], "ensemble_add_sub_config_1"
-        )  # model config name
-        self.assertEqual(summary_table._rows[0][1], "0, 0")  # max batch size
-        self.assertEqual(
-            summary_table._rows[0][2], "(Disabled, Disabled)"
-        )  # dynamic batching
-        self.assertEqual(summary_table._rows[0][3], "1:GPU, 2:GPU")  # instance count
-        self.assertEqual(summary_table._rows[0][4], "3.655")  # p99 latency
-        self.assertEqual(summary_table._rows[0][5], "2865.88")  # throughput
-
-    def test_ensemble_detailed(self):
-        """
-        Ensures the detailed report sentence is accurate for a basic ensemble model (loaded from a checkpoint)
-        """
-        self._init_managers(models="ensemble_add_sub_config_1", subcommand="report")
-
-        self.report_manager._add_detailed_report_data()
-        self.report_manager._build_detailed_table("ensemble_add_sub_config_1")
-        detailed_sentence = self.report_manager._build_detailed_info(
-            "ensemble_add_sub_config_1"
-        )
-
-        expected_detailed_sentence = (
-            "<strong>ensemble_add_sub_config_1</strong> is comprised of the "
-            "following composing models:<LI> <strong>add_config_0</strong>: "
-            "1 GPU instance with a max batch size of 0 on platform python </LI><LI> "
-            "<strong>sub_config_1</strong>: 2 GPU instances with a max batch size of 0 "
-            "on platform python </LI><br>5 measurement(s) were obtained for the model config "
-            "on GPU(s)  with total memory 0 GB."
-        )
-
-        self.assertEqual(detailed_sentence, expected_detailed_sentence)
-
-    def tearDown(self):
-        patch.stopall()
-        shutil.rmtree(f"{ROOT_DIR}/reports")
-        shutil.rmtree(f"{ROOT_DIR}/plots")
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_file_writer.py b/src/c++/perf_analyzer/tests/test_file_writer.py
deleted file mode 100755
index f86ec393a..000000000
--- a/src/c++/perf_analyzer/tests/test_file_writer.py
+++ /dev/null
@@ -1,67 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.output.file_writer import FileWriter
-
-from .common import test_result_collector as trc
-from .mocks.mock_io import MockIOMethods
-
-TEST_FILENAME = "test_filename"
-
-
-class TestFileWriterMethods(trc.TestResultCollector):
-    def setUp(self):
-        self._mock_path = "model_analyzer.output.file_writer"
-        self.io_mock = MockIOMethods(mock_paths=["model_analyzer.output.file_writer"])
-        self.io_mock.start()
-
-    def test_write(self):
-        # Create and use writer
-        writer = FileWriter(filename=TEST_FILENAME)
-
-        # Check for exception on open and write
-        self.io_mock.raise_exception_on_open()
-        err_str = "Expected TritonModelAnalyzerException on malformed input."
-        with self.assertRaises(TritonModelAnalyzerException, msg=err_str):
-            writer.write("test")
-        self.io_mock.reset()
-
-        self.io_mock.raise_exception_on_write()
-        err_str = "Expected TritonModelAnalyzerException on malformed input."
-        with self.assertRaises(TritonModelAnalyzerException, msg=err_str):
-            writer.write("test")
-        self.io_mock.reset()
-
-        # Check mock call on successful write
-        writer.write("test")
-        self.io_mock.assert_write_called_with_args(self._mock_path, "test")
-
-        # Perform checks for stdout
-        writer = FileWriter()
-        writer.write("test")
-        self.io_mock.assert_print_called_with_args(self._mock_path, "test")
-
-    def tearDown(self):
-        patch.stopall()
-        self.io_mock.stop()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_html_report.py b/src/c++/perf_analyzer/tests/test_html_report.py
deleted file mode 100755
index 9473ede9c..000000000
--- a/src/c++/perf_analyzer/tests/test_html_report.py
+++ /dev/null
@@ -1,167 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import base64
-import unittest
-from unittest.mock import mock_open, patch
-
-from model_analyzer.reports.html_report import HTMLReport
-from model_analyzer.result.result_table import ResultTable
-
-from .common import test_result_collector as trc
-
-
-class TestHTMLReportMethods(trc.TestResultCollector):
-    """
-    Tests the methods of the HTMLReport class
-    """
-
-    def setUp(self):
-        self.maxDiff = None
-        self.report = HTMLReport()
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_add_title(self):
-        self.report.add_title("Test HTML Report")
-        expected_report_body = "<html><head><style></style></head><body><center><h1>Test HTML Report</h1></center></body></html>"
-        self.assertEqual(self.report.document(), expected_report_body)
-
-    def test_add_subheading(self):
-        # Add one subheading
-        self.report.add_subheading("Throughput vs. Latency")
-        expected_report_body = "<html><head><style></style></head><body><h3>Throughput vs. Latency</h3></body></html>"
-        self.assertEqual(self.report.document(), expected_report_body)
-
-        # Add another subheading
-        self.report.add_subheading("GPU Memory vs. Latency")
-        expected_report_body = (
-            "<html><head><style></style></head><body><h3>Throughput vs. Latency</h3>"
-            "<h3>GPU Memory vs. Latency</h3></body></html>"
-        )
-        self.assertEqual(self.report.document(), expected_report_body)
-
-    def test_add_image(self):
-        with patch(
-            "model_analyzer.reports.html_report.open",
-            mock_open(read_data=bytes(">:(".encode("ascii"))),
-        ):
-            self.report.add_images(["test_image_file"], ["test_caption"])
-        img_content = base64.b64encode(bytes(">:(".encode("ascii"))).decode("ascii")
-        expected_report_body = (
-            f'<html><head><style></style></head><body><center><div><div class="image" style="float:center;width:100%"><img src="data:image/png;base64,{img_content}"'
-            ' style="width:100%"><center><div style="font-weight:bold;font-size:12;padding-bottom:20px">'
-            "test_caption</div></center></div></div></center></body></html>"
-        )
-        self.assertEqual(self.report.document(), expected_report_body)
-
-    def test_add_paragraph(self):
-        test_paragraph = (
-            "This is a test paragraph with a lot to say."
-            " There is more than one line in this paragraph."
-        )
-
-        # Default font size
-        self.report.add_paragraph(test_paragraph)
-        expected_report_body = (
-            '<html><head><style></style></head><body><div style="font-size:14">'
-            f"<p>{test_paragraph}</p></div></body></html>"
-        )
-        self.assertEqual(self.report.document(), expected_report_body)
-
-        # custom font size
-        self.report.add_paragraph(test_paragraph, font_size=20)
-        expected_report_body = (
-            '<html><head><style></style></head><body><div style="font-size:14">'
-            f'<p>{test_paragraph}</p></div><div style="font-size:20">'
-            f"<p>{test_paragraph}</p></div></body></html>"
-        )
-        self.assertEqual(self.report.document(), expected_report_body)
-
-    def test_add_table(self):
-        result_table = ResultTable(["header1", "header2"])
-        # Try empty table
-        self.report.add_table(table=result_table)
-        table_style = "border: 1px solid black;border-collapse: collapse;text-align: center;width: 80%;padding: 5px 10px;font-size: 11pt"
-        expected_report_body = (
-            "<html><head><style></style></head><body>"
-            f'<center><table style="{table_style}">'
-            "<tr>"
-            f'<th style="{table_style}">header1</th>'
-            f'<th style="{table_style}">header2</th>'
-            "</tr>"
-            "</table></center>"
-            "</body></html>"
-        )
-        self.assertEqual(self.report.document(), expected_report_body)
-
-        # Fill table
-        for i in range(2):
-            result_table.insert_row_by_index([f"value{i}1", f"value{i}2"])
-
-        # Table has 5 rows
-        self.report.add_table(table=result_table)
-        expected_report_body = (
-            "<html><head><style></style></head><body>"
-            f'<center><table style="{table_style}">'
-            "<tr>"
-            f'<th style="{table_style}">header1</th>'
-            f'<th style="{table_style}">header2</th>'
-            "</tr>"
-            "</table></center>"
-            f'<center><table style="{table_style}">'
-            "<tr>"
-            f'<th style="{table_style}">header1</th>'
-            f'<th style="{table_style}">header2</th>'
-            "</tr>"
-            "<tr>"
-            f'<td style="{table_style}">value01</td>'
-            f'<td style="{table_style}">value02</td>'
-            "</tr>"
-            "<tr>"
-            f'<td style="{table_style}">value11</td>'
-            f'<td style="{table_style}">value12</td>'
-            "</tr>"
-            "</table></center>"
-            "</body></html>"
-        )
-
-        self.assertEqual(self.report.document(), expected_report_body)
-
-    def test_write_report(self):
-        self.report.add_title("Test HTML Report")
-        self.report.add_subheading("Throughput vs. Latency")
-        test_paragraph = (
-            "This is a test paragraph with a lot to say."
-            " There is more than one line in this paragraph."
-        )
-        self.report.add_paragraph(test_paragraph, font_size=14)
-
-        expected_report_body = (
-            "<html><head><style></style></head><body><center><h1>Test HTML Report</h1></center><h3>Throughput vs. Latency</h3>"
-            f'<div style="font-size:14"><p>{test_paragraph}</p></div>'
-            "</body></html>"
-        )
-
-        with patch("builtins.open", mock_open()) as mocked_file:
-            self.report.write_report("test_report_filename")
-        mocked_file.assert_called_once_with(f"test_report_filename", "w")
-        mocked_file().write.assert_called_once_with(f"{expected_report_body}")
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_logger.py b/src/c++/perf_analyzer/tests/test_logger.py
deleted file mode 100755
index fec7c1e88..000000000
--- a/src/c++/perf_analyzer/tests/test_logger.py
+++ /dev/null
@@ -1,175 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-# This test exercises all of the cli options.
-# It does a basic test of the cli and help message.
-# Then the rest of the testing uses an OptionStruct, which holds all
-# of the data necessary to test a command, and feeds that to the
-# CLI parser. The result of the CLI parsing is compared against the
-# expected value for the CLI. Default values are also verified as well as
-# values that are expected to cause failures.
-
-import logging
-import re
-import sys
-import unittest
-from contextlib import contextmanager
-from io import StringIO
-from unittest.mock import patch
-
-from model_analyzer.constants import LOGGER_NAME
-from model_analyzer.log_formatter import setup_logging
-
-from .common import test_result_collector as trc
-
-logger = logging.getLogger(LOGGER_NAME)
-
-
-@contextmanager
-def captured_output():
-    new_out, new_err = StringIO(), StringIO()
-    old_out, old_err = sys.stdout, sys.stderr
-    try:
-        sys.stdout, sys.stderr = new_out, new_err
-        yield sys.stdout, sys.stderr
-    finally:
-        sys.stdout, sys.stderr = old_out, old_err
-
-
-class TestLogger(trc.TestResultCollector):
-    def tearDown(self):
-        patch.stopall()
-
-    def test_info_normal(self):
-        """Test expected format of logger.info"""
-        with captured_output() as (out, err):
-            setup_logging(verbose=False, quiet=False)
-            logger.info("ABC")
-            expected_output = "[Model Analyzer] ABC"
-            output = out.getvalue().strip()
-            self.assertEqual(output, expected_output)
-
-    def test_info_quiet(self):
-        """
-        Test expected format of logger.info in quiet mode
-        (Nothing should be printed)
-        """
-        with captured_output() as (out, err):
-            setup_logging(verbose=False, quiet=True)
-            logger.info("ABC")
-            expected_output = ""
-            output = out.getvalue().strip()
-            self.assertEqual(output, expected_output)
-
-    def test_info_verbose(self):
-        """
-        Test expected format of logger.info in verbose mode
-        (Extra date info should be added to the start of the line)
-        """
-        with captured_output() as (out, err):
-            setup_logging(verbose=True, quiet=False)
-            logger.info("ABC")
-            output = out.getvalue().strip()
-            self.assertTrue(re.match("^\d\d:\d\d:\d\d \[Model Analyzer\] ABC$", output))
-
-    def test_debug_normal(self):
-        """
-        Test expected format of logger.debug
-        (Nothing should be printed since we aren't in verbose mode)
-        """
-        with captured_output() as (out, err):
-            setup_logging(verbose=False, quiet=False)
-            logger.debug("ABC")
-            expected_output = ""
-            output = out.getvalue().strip()
-            self.assertEqual(output, expected_output)
-
-    def test_debug_quiet(self):
-        """
-        Test expected format of logger.debug in quiet mode
-        (Nothing should be printed since we aren't in verbose mode)
-        """
-        with captured_output() as (out, err):
-            setup_logging(verbose=False, quiet=True)
-            logger.debug("ABC")
-            expected_output = ""
-            output = out.getvalue().strip()
-            self.assertEqual(output, expected_output)
-
-    def test_debug_verbose(self):
-        """
-        Test expected format of logger.debug in verbose mode
-        (Extra date info should be added to the start of the line)
-        """
-        with captured_output() as (out, err):
-            setup_logging(verbose=True, quiet=False)
-            logger.debug("ABC")
-            output = out.getvalue().strip()
-            self.assertTrue(
-                re.match("^\d\d:\d\d:\d\d \[Model Analyzer\] DEBUG: ABC$", output)
-            )
-
-    def test_error_normal(self):
-        """
-        Test expected format of logger.error
-        """
-        with captured_output() as (out, err):
-            setup_logging(verbose=False, quiet=False)
-            logger.error("ABC")
-            expected_output = "[Model Analyzer] ERROR: ABC"
-            output = out.getvalue().strip()
-            self.assertEqual(output, expected_output)
-
-    def test_error_quiet(self):
-        """
-        Test expected format of logger.error in quiet mode
-        (It will be printed despite verbose mode)
-        """
-        with captured_output() as (out, err):
-            setup_logging(verbose=False, quiet=True)
-            logger.error("ABC")
-            expected_output = "[Model Analyzer] ERROR: ABC"
-            output = out.getvalue().strip()
-            self.assertEqual(output, expected_output)
-
-    def test_error_verbose(self):
-        """
-        Test expected format of logger.error in verbose mode
-        (Extra date info should be added to the start of the line)
-        """
-        with captured_output() as (out, err):
-            setup_logging(verbose=True, quiet=False)
-            logger.error("ABC")
-            output = out.getvalue().strip()
-            self.assertTrue(
-                re.match("^\d\d:\d\d:\d\d \[Model Analyzer\] ERROR: ABC$", output)
-            )
-
-    def test_quiet_trumps_verbose(self):
-        """
-        Test that calling setup_logging with both verbose and quiet true
-        will result in quiet mode
-        """
-        with captured_output() as (out, err):
-            setup_logging(verbose=True, quiet=True)
-            logger.info("ABC")
-            expected_output = ""
-            output = out.getvalue().strip()
-            self.assertEqual(output, expected_output)
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_config.py b/src/c++/perf_analyzer/tests/test_model_config.py
deleted file mode 100755
index 5371107dc..000000000
--- a/src/c++/perf_analyzer/tests/test_model_config.py
+++ /dev/null
@@ -1,472 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os
-import unittest
-from unittest.mock import MagicMock, mock_open, patch
-
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.triton.model.model_config import ModelConfig
-
-from .common import test_result_collector as trc
-from .mocks.mock_model_config import MockModelConfig
-
-
-class TestModelConfig(trc.TestResultCollector):
-    def setUp(self):
-        self._model_config = {
-            "name": "classification_chestxray_v1",
-            "platform": "tensorflow_graphdef",
-            "max_batch_size": 32,
-            "input": [
-                {
-                    "name": "NV_MODEL_INPUT",
-                    "data_type": "TYPE_FP32",
-                    "format": "FORMAT_NHWC",
-                    "dims": ["256", "256", "3"],
-                }
-            ],
-            "output": [
-                {
-                    "name": "NV_MODEL_OUTPUT",
-                    "data_type": "TYPE_FP32",
-                    "dims": ["15"],
-                    "label_filename": "chestxray_labels.txt",
-                }
-            ],
-            "instance_group": [{"count": 1, "kind": "KIND_GPU"}],
-        }
-
-        # Equivalent protobuf for the model config above.
-        self._model_config_protobuf = """
-name: "classification_chestxray_v1"
-platform: "tensorflow_graphdef"
-max_batch_size: 32
-input [
-  {
-    name: "NV_MODEL_INPUT"
-    data_type: TYPE_FP32
-    format: FORMAT_NHWC
-    dims: [256, 256, 3]
-  }
-]
-output [
-  {
-    name: "NV_MODEL_OUTPUT"
-    data_type: TYPE_FP32
-    dims: [15]
-    label_filename: "chestxray_labels.txt"
-  }
-]
-instance_group [
-  {
-    kind: KIND_GPU
-    count: 1
-  }
-]
-"""
-
-        # Ensemble model config
-        self._ensemble_model_config_protobuf = """
-    name: "ensemble_python_resnet50"
-platform: "ensemble"
-max_batch_size: 256
-input [
-  {
-    name: "INPUT"
-    data_type: TYPE_UINT8
-    dims: [ -1 ]
-  }
-]
-output [
-  {
-    name: "OUTPUT"
-    data_type: TYPE_FP32
-    dims: [ 1000 ]
-  }
-]
-ensemble_scheduling {
-  step [
-    {
-      model_name: "preprocess"
-      model_version: -1
-      input_map {
-        key: "INPUT_0"
-        value: "INPUT"
-      }
-      output_map {
-        key: "OUTPUT_0"
-        value: "preprocessed_image"
-      }
-    },
-    {
-      model_name: "resnet50_trt"
-      model_version: -1
-      input_map {
-        key: "input"
-        value: "preprocessed_image"
-      }
-      output_map {
-        key: "output"
-        value: "OUTPUT"
-      }
-    }
-  ]
-}
-"""
-
-        self._ensemble_model_config_protobuf_no_scheduling = """
-    name: "ensemble_python_resnet50"
-platform: "ensemble"
-max_batch_size: 256
-input [
-  {
-    name: "INPUT"
-    data_type: TYPE_UINT8
-    dims: [ -1 ]
-  }
-]
-output [
-  {
-    name: "OUTPUT"
-    data_type: TYPE_FP32
-    dims: [ 1000 ]
-  }
-]
-"""
-
-        self._ensemble_model_config_protobuf_no_step = """
-    name: "ensemble_python_resnet50"
-platform: "ensemble"
-max_batch_size: 256
-input [
-  {
-    name: "INPUT"
-    data_type: TYPE_UINT8
-    dims: [ -1 ]
-  }
-]
-output [
-  {
-    name: "OUTPUT"
-    data_type: TYPE_FP32
-    dims: [ 1000 ]
-  }
-]
-ensemble_scheduling {
-}
-"""
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_create_from_file(self):
-        test_protobuf = self._model_config_protobuf
-        mock_model_config = MockModelConfig(test_protobuf)
-        mock_model_config.start()
-        model_config = ModelConfig._create_from_file("/path/to/model_config")
-        self.assertEqual(model_config.get_config(), self._model_config)
-        mock_model_config.stop()
-
-    def test_create_from_dict(self):
-        model_config = ModelConfig.create_from_dictionary(self._model_config)
-        self.assertEqual(model_config.get_config(), self._model_config)
-
-        new_config = {"instance_group": [{"count": 2, "kind": "KIND_CPU"}]}
-        model_config.set_config(new_config)
-        self.assertEqual(model_config.get_config(), new_config)
-
-    def test_write_config_file(self):
-        model_config = ModelConfig.create_from_dictionary(self._model_config)
-        model_output_path = os.path.abspath("./model_config")
-
-        mock_model_config = MockModelConfig()
-        mock_model_config.start()
-        # Write the model config to output
-        with patch(
-            "model_analyzer.triton.model.model_config.open", mock_open()
-        ) as mocked_file:
-            with patch(
-                "model_analyzer.triton.model.model_config.copytree", MagicMock()
-            ):
-                model_config.write_config_to_file(model_output_path, "/mock/path", None)
-            content = mocked_file().write.call_args.args[0]
-        mock_model_config.stop()
-
-        mock_model_config = MockModelConfig(content)
-        mock_model_config.start()
-        model_config_from_file = ModelConfig._create_from_file(model_output_path)
-        self.assertEqual(model_config_from_file.get_config(), self._model_config)
-        mock_model_config.stop()
-
-        # output path doesn't exist
-        with patch(
-            "model_analyzer.triton.model.model_config.os.path.exists",
-            MagicMock(return_value=False),
-        ):
-            with self.assertRaises(TritonModelAnalyzerException):
-                ModelConfig._create_from_file(model_output_path)
-
-        # output path is a file
-        with patch(
-            "model_analyzer.triton.model.model_config.os.path.isfile",
-            MagicMock(return_value=True),
-        ):
-            with self.assertRaises(TritonModelAnalyzerException):
-                ModelConfig._create_from_file(model_output_path)
-
-    @patch(
-        "model_analyzer.triton.model.model_config.os.listdir",
-        MagicMock(return_value=["1", "config.pbtxt", "output0_labels.txt"]),
-    )
-    @patch("model_analyzer.triton.model.model_config.copytree")
-    @patch("model_analyzer.triton.model.model_config.os.symlink")
-    def test_write_config_to_file_with_relative_path(self, mock_os_symlink, *args):
-        """
-        Tests that the call to os.symlink() within write_config_to_file() uses
-        a valid relative path when user uses a relative path with the
-        `--output-model-repository-path` option
-        """
-
-        model_config = ModelConfig.create_from_dictionary(self._model_config)
-
-        model_path = "./output_model_repository/model_config_1"
-        src_model_path = "/tmp/src_model_repository/model"
-        last_model_path = "./output_model_repository/model_config_0"
-
-        mock_model_config = MockModelConfig()
-        mock_model_config.start()
-        model_config.write_config_to_file(model_path, src_model_path, last_model_path)
-        mock_model_config.stop()
-
-        mock_os_symlink.assert_any_call(
-            "../model_config_0/1", "./output_model_repository/model_config_1/1"
-        )
-        mock_os_symlink.assert_any_call(
-            "../model_config_0/output0_labels.txt",
-            "./output_model_repository/model_config_1/output0_labels.txt",
-        )
-
-    def test_instance_group_string(self):
-        """Test out all corner cases of instance_group_string()"""
-
-        def _test_helper(config_dict, expected_result, gpu_count=None):
-            model_config = ModelConfig.create_from_dictionary(config_dict)
-            if gpu_count:
-                instance_group_str = model_config.instance_group_string(gpu_count)
-            else:
-                instance_group_str = model_config.instance_group_string()
-            self.assertEqual(instance_group_str, expected_result)
-
-        # No instance group info in model_config_dict:
-        #  - default to 1 per GPU
-        model_config_dict = {}
-        _test_helper(model_config_dict, "1:GPU", gpu_count=1)
-
-        # No instance group info in model_config_dict:
-        #  - 1 per GPU -- if 2 gpus then 2 total
-        model_config_dict = {}
-        _test_helper(model_config_dict, "2:GPU", gpu_count=2)
-
-        # No instance group info in model_config_dict:
-        #  - default to 1 on CPU if cuda not available
-        model_config_dict = {}
-        with patch("numba.cuda.is_available", MagicMock(return_value=False)):
-            _test_helper(model_config_dict, "1:CPU", gpu_count=5)
-
-        # 2 per GPU, 3 gpus in the system = 6 total
-        model_config_dict = {
-            "instance_group": [
-                {
-                    "count": 2,
-                    "kind": "KIND_GPU",
-                }
-            ]
-        }
-        _test_helper(model_config_dict, "6:GPU", gpu_count=3)
-
-        # 1 on GPU0 only = 1 total despite 2 GPUs in the system
-        model_config_dict = {
-            "instance_group": [{"count": 1, "kind": "KIND_GPU", "gpus": [0]}]
-        }
-        _test_helper(model_config_dict, "1:GPU", gpu_count=2)
-
-        # 1 on ALL gpus + 2 each on [1 and 3] + 3 more on CPUs
-        # with 4 GPUs in the system:
-        #   8 on GPU and 3 on CPU
-        model_config_dict = {
-            "instance_group": [
-                {"count": 1, "kind": "KIND_GPU"},
-                {"count": 2, "kind": "KIND_GPU", "gpus": [1, 3]},
-                {"count": 3, "kind": "KIND_CPU"},
-            ]
-        }
-        _test_helper(model_config_dict, "8:GPU + 3:CPU", gpu_count=4)
-
-    def test_instance_group_count(self):
-        """Test out all corner cases of instance_group_count()"""
-
-        def _test_helper(config_dict, expected_result, gpu_count=None):
-            model_config = ModelConfig.create_from_dictionary(config_dict)
-            instance_group_count = model_config.instance_group_count(gpu_count)
-            self.assertEqual(instance_group_count, expected_result)
-
-        # No instance group info in model_config_dict:
-        #  - default to 1 per GPU
-        model_config_dict = {}
-        _test_helper(model_config_dict, 1, gpu_count=1)
-
-        # No instance group info in model_config_dict:
-        #  - 1 per GPU -- if 2 gpus then 2 total
-        model_config_dict = {}
-        _test_helper(model_config_dict, 2, gpu_count=2)
-
-        # 2 per GPU, 3 gpus in the system = 6 total
-        model_config_dict = {
-            "instance_group": [
-                {
-                    "count": 2,
-                    "kind": "KIND_GPU",
-                }
-            ]
-        }
-        _test_helper(model_config_dict, 6, gpu_count=3)
-
-        # 1 on ALL gpus + 2 each on [1 and 3] + 3 more on CPUs
-        # with 4 GPUs in the system:
-        #   8 on GPU and 3 on CPU
-        model_config_dict = {
-            "instance_group": [
-                {"count": 1, "kind": "KIND_GPU"},
-                {"count": 2, "kind": "KIND_GPU", "gpus": [1, 3]},
-                {"count": 3, "kind": "KIND_CPU"},
-            ]
-        }
-        _test_helper(model_config_dict, 11, gpu_count=4)
-
-    def test_is_ensemble(self):
-        """Test that we recognize if the platform is ensemble"""
-
-        tensor_protobuf = self._model_config_protobuf
-        mock_model_config = MockModelConfig(tensor_protobuf)
-        mock_model_config.start()
-        model_config = ModelConfig._create_from_file("/path/to/model_config")
-        self.assertFalse(model_config.is_ensemble())
-        mock_model_config.stop()
-
-        ensemble_protobuf = self._ensemble_model_config_protobuf
-        mock_model_config = MockModelConfig(ensemble_protobuf)
-        mock_model_config.start()
-        model_config = ModelConfig._create_from_file("/path/to/model_config")
-        self.assertTrue(model_config.is_ensemble())
-        mock_model_config.stop()
-
-    def test_ensemble_composing_models(self):
-        """Test that we can extract an ensembles composing model names"""
-
-        ensemble_protobuf = self._ensemble_model_config_protobuf
-        mock_model_config = MockModelConfig(ensemble_protobuf)
-        mock_model_config.start()
-        model_config = ModelConfig._create_from_file("/path/to/model_config")
-
-        self.assertEqual(
-            model_config.get_ensemble_composing_models(), ["preprocess", "resnet50_trt"]
-        )
-        mock_model_config.stop()
-
-    def test_set_composing_model_name(self):
-        """Test setting a variant name for an ensemble's composing model"""
-
-        ensemble_protobuf = self._ensemble_model_config_protobuf
-        mock_model_config = MockModelConfig(ensemble_protobuf)
-        mock_model_config.start()
-        model_config = ModelConfig._create_from_file("/path/to/model_config")
-
-        model_config.set_composing_model_variant_name(
-            composing_model_name="preprocess", variant_name="preprocess_config_0"
-        )
-        model_config.set_composing_model_variant_name(
-            composing_model_name="resnet50_trt", variant_name="resnet50_trt_config_1"
-        )
-
-        self.assertEqual(
-            model_config.get_ensemble_composing_models(),
-            ["preprocess_config_0", "resnet50_trt_config_1"],
-        )
-        mock_model_config.stop()
-
-    def test_badly_formed_ensemble_configs(self):
-        """Test cases where protobuf is incorrect for an ensemble platform"""
-
-        # Tests where platform is not ensemble
-        ensemble_protobuf = self._model_config_protobuf
-        mock_model_config = MockModelConfig(ensemble_protobuf)
-        mock_model_config.start()
-        model_config = ModelConfig._create_from_file("/path/to/model_config")
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            model_config.get_ensemble_composing_models()
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            model_config.set_composing_model_variant_name(
-                composing_model_name="preprocess", variant_name="preprocess_config_0"
-            )
-
-        mock_model_config.stop()
-
-        # Tests where platform is ensemble, but scheduling section doesn't exist
-        ensemble_protobuf = self._ensemble_model_config_protobuf_no_scheduling
-        mock_model_config = MockModelConfig(ensemble_protobuf)
-        mock_model_config.start()
-        model_config = ModelConfig._create_from_file("/path/to/model_config")
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            model_config.get_ensemble_composing_models()
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            model_config.set_composing_model_variant_name(
-                composing_model_name="preprocess", variant_name="preprocess_config_0"
-            )
-
-        mock_model_config.stop()
-
-        # Tests where platform is ensemble, scheduling exists, but step is omitted
-        ensemble_protobuf = self._ensemble_model_config_protobuf_no_step
-        mock_model_config = MockModelConfig(ensemble_protobuf)
-        mock_model_config.start()
-        model_config = ModelConfig._create_from_file("/path/to/model_config")
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            model_config.get_ensemble_composing_models()
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            model_config.set_composing_model_variant_name(
-                composing_model_name="preprocess", variant_name="preprocess_config_0"
-            )
-
-        mock_model_config.stop()
-
-    def test_set_model_name(self):
-        """
-        Test that we can set the model name of a config
-        """
-        model_config = ModelConfig.create_from_dictionary(self._model_config)
-        model_config.set_model_name("new_model_name")
-        self.assertEqual(model_config.get_field("name"), "new_model_name")
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_config_generator.py b/src/c++/perf_analyzer/tests/test_model_config_generator.py
deleted file mode 100755
index 3717c04be..000000000
--- a/src/c++/perf_analyzer/tests/test_model_config_generator.py
+++ /dev/null
@@ -1,927 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.config.generate.base_model_config_generator import (
-    BaseModelConfigGenerator,
-)
-from model_analyzer.config.generate.model_config_generator_factory import (
-    ModelConfigGeneratorFactory,
-)
-from model_analyzer.config.generate.model_profile_spec import ModelProfileSpec
-from model_analyzer.config.generate.model_variant_name_manager import (
-    ModelVariantNameManager,
-)
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.triton.model.model_config import ModelConfig
-
-from .common import test_result_collector as trc
-from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
-from .mocks.mock_model_config import MockModelConfig
-from .mocks.mock_os import MockOSMethods
-
-
-class TestModelConfigGenerator(trc.TestResultCollector):
-    def __init__(self, methodname):
-        super().__init__(methodname)
-        self._fake_throughput = 1
-
-    def test_direct_no_params(self):
-        """
-        Test direct modes with no model_config_parameters specified
-
-        It will just sweep instance count and max_batch_size (with dynamic batching on),
-        and default config (empty dict) will be included
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 1, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 32, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 64, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 128, 'instance_group': [{'count': 5, 'kind': 'KIND_GPU'}]}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_direct_no_params_search_disable(self):
-        """
-        Test direct mode with no model_config_parameters specified and run_search disabled
-
-        This will just return a single empty config, since there are no parameters to combine
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_disable: True
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [{'max_batch_size': 8}]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_direct_yes_params_search_disable(self):
-        """
-        Test direct modes with model_config_parameters specified and run_search disabled
-
-        This will just combine all model_config_parameters
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_instance_count: 16
-            run_config_search_disable: True
-            profile_models:
-                test_model:
-                    model_config_parameters:
-                        max_batch_size: [1,4,16]
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: [1,2]
-            """)
-
-        expected_configs = [
-            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
-            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
-            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 16},
-            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
-            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
-            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 16}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_run_config_search_options(self):
-        """
-        Test that ModelConfigGenerator will honor run_config_search_max_instance_count
-        and run_config_search_max_model_batch_size and run_config_search_min_model_batch_size
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_instance_count: 3
-            run_config_search_min_model_batch_size: 2
-            run_config_search_max_model_batch_size: 16
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_run_config_search_min_instance_counts(self):
-        """
-        Test that ModelConfigGenerator will honor run_config_search_min_instance_count
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_min_instance_count: 2
-            run_config_search_max_instance_count: 3
-            run_config_search_min_model_batch_size: 2
-            run_config_search_max_model_batch_size: 8
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_non_power_of_two_max_batch_size(self):
-        """
-        Test that ModelConfigGenerator will correctly sweep max_batch_size with
-        input values that aren't a power of 2
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_instance_count: 2
-            run_config_search_min_model_batch_size: 3
-            run_config_search_max_model_batch_size: 15
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'dynamic_batching': {}, 'max_batch_size': 3, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 3, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 6, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 6, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 12, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 12, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_direct_yes_params_specified(self):
-        """
-        Test direct mode with model_config_parameters specified
-
-        It will combine all legal combinations of config values, and
-        default config (None) will be included
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_instance_count: 16
-            run_config_search_disable: False
-            profile_models:
-                test_model:
-                    model_config_parameters:
-                        max_batch_size: [1,4,16]
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: [1,2]
-            """)
-
-        expected_configs = [
-            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
-            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
-            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 16},
-            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
-            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
-            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 16}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_direct_cpu_only(self):
-        """
-        Test direct mode with cpu_only=true
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_instance_count: 2
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 16
-            profile_models:
-                - my-model:
-                    cpu_only: True
-            """)
-
-        expected_configs = [
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_CPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_CPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 1, 'kind': 'KIND_CPU'}]},
-            {'dynamic_batching': {}, 'max_batch_size': 16, 'instance_group': [{'count': 2, 'kind': 'KIND_CPU'}]}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_direct_max_batch_size_0(self):
-        """
-        Test direct mode with the the default config stating max_batch_size=0
-
-        max_batch_size and dynamic_batching should not be part of the resulting configs
-        """
-
-        # yapf: disable
-        protobuf = """
-            platform: "fake_platform"
-            max_batch_size: 0
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 1
-            }
-            ]
-            """
-
-        yaml_str = ("""
-            run_config_search_max_instance_count: 4
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'platform': "fake_platform", 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
-
-    def test_direct_max_batch_size_unspecified(self):
-        """
-        Test direct mode with the the default config not specifying max_batch_size
-
-        max_batch_size and dynamic_batching should not be part of the resulting configs
-        """
-
-        # yapf: disable
-        protobuf = """
-            platform: "fake_platform"
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 1
-            }
-            ]
-            """
-
-        yaml_str = ("""
-            run_config_search_max_instance_count: 4
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'platform': "fake_platform", 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
-
-    def test_direct_sequence_batching(self):
-        """
-        Test direct mode with the the default config specifying sequence_batching
-
-        dynamic_batching should not be part of the resulting configs
-        """
-
-        # yapf: disable
-        protobuf = """
-            platform: "fake_platform"
-            max_batch_size: 4
-            sequence_batching {}
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 1
-            }
-            ]
-            """
-
-        yaml_str = ("""
-            run_config_search_max_instance_count: 4
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 16
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'platform': "fake_platform", 'max_batch_size':  8, 'sequence_batching': {}, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'max_batch_size':  8, 'sequence_batching': {}, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'max_batch_size':  8, 'sequence_batching': {}, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'max_batch_size':  8, 'sequence_batching': {}, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'max_batch_size': 16, 'sequence_batching': {}, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'max_batch_size': 16, 'sequence_batching': {}, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'max_batch_size': 16, 'sequence_batching': {}, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'platform': "fake_platform", 'max_batch_size': 16, 'sequence_batching': {}, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}]}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
-
-    def test_direct_nonempty_default_config(self):
-        """
-        Test direct mode with the the default config containing some values
-
-        It will keep values that aren't part of the search, and will overwrite
-        any values that are part of the search
-        """
-
-        # yapf: disable
-        protobuf = """
-            platform: "fake_platform"
-            max_batch_size: 4
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 1
-            }
-            ]
-            """
-
-        yaml_str = ("""
-            run_config_search_max_instance_count: 4
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'platform': "fake_platform", 'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'platform': "fake_platform", 'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'platform': "fake_platform", 'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'platform': "fake_platform", 'max_batch_size': 8, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}],'dynamic_batching': {}}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
-
-    def test_remote_yes_params_specified(self):
-        """
-        Test remote mode with model_config_parameters specified
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            triton_launch_mode: remote
-            run_config_search_max_instance_count: 16
-            profile_models:
-                test_model:
-                    model_config_parameters:
-                        max_batch_size: [1,4,16]
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: [1,2]
-            """)
-
-        expected_configs = [
-            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
-            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
-            {'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}], 'max_batch_size': 16},
-            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 1},
-            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 4},
-            {'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}], 'max_batch_size': 16}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_remote_no_params_specified(self):
-        """
-        Test remote mode with no model_config_parameters specified
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            triton_launch_mode: remote
-            run_config_search_max_instance_count: 4
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 8, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}],'dynamic_batching': {}}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs)
-
-    def test_search_subparameter(self):
-        """
-        Test that if a subparameter is swept, that it will not overwrite other subparameters
-
-        Param2 should exist in all results despite the fact that param1 is the one being swept.
-        Also, gpu_execution_accelerator (a nested subproperty) should still exist despite a
-        sibling property (cpu_execution_accelerator) being overwritten
-        """
-
-        # yapf: disable
-        protobuf = """
-            max_batch_size: 4
-            instance_group [
-            {
-                kind: KIND_GPU
-                count: 1
-            }
-            ]
-            parameters [
-                {
-                    key: "param1"
-                    value: { string_value: "param1_value" }
-                },
-                {
-                    key: "param2"
-                    value: { string_value: "param2_value" }
-                }
-            ]
-            optimization { execution_accelerators {
-                cpu_execution_accelerator : [ {
-                    name : "fake_cpu_accelerator"
-                }]
-                gpu_execution_accelerator : [ {
-                    name : "fake_gpu_accelerator"
-                }]
-            }}
-            """
-
-        yaml_str = ("""
-            profile_models:
-                my-model:
-                    model_config_parameters:
-                        optimization:
-                            execution_accelerators:
-                                cpu_execution_accelerator:
-                                - name: "new_cpu_accelerator"
-                        parameters:
-                            param1:
-                                string_value: ["foo", "bar"]
-            """)
-
-        expected_configs = [
-            {
-                'optimization':
-                {
-                    'execution_accelerators':
-                    {
-                        'cpu_execution_accelerator': [{'name': 'new_cpu_accelerator'}],
-                        'gpu_execution_accelerator': [{'name': 'fake_gpu_accelerator'}]
-                    }
-                },
-                'parameters':
-                {
-                    'param1': {'string_value': 'foo'},
-                    'param2': {'string_value': 'param2_value'}
-                },
-                'max_batch_size': 4,
-                'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]
-            },
-            {
-                'optimization':
-                {
-                    'execution_accelerators':
-                    {
-                        'cpu_execution_accelerator': [{'name': 'new_cpu_accelerator'}],
-                        'gpu_execution_accelerator': [{'name': 'fake_gpu_accelerator'}]
-                    }
-                },
-                'parameters':
-                {
-                    'param1': {'string_value': 'bar'},
-                    'param2': {'string_value': 'param2_value'}
-                },
-                'max_batch_size': 4,
-                'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]
-            },
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
-
-    def test_search_dynamic_batching_subparameter(self):
-        """
-        Test that if dynamic batching was already on with subparameters, they will not be overwritten in automatic search
-
-        Normally automatic search just turns on dynamic_batching, but in this case dynamic_batching is already
-        on in the user's default model config with a value for max_queue_delay_microseconds. That value should remain
-        unchanged in all model configs generated by the manual search
-        """
-
-        # yapf: disable
-        protobuf = """
-            max_batch_size: 4
-            instance_group [
-            {
-                kind: KIND_GPU
-                count: 1
-            }
-            ]
-            dynamic_batching {
-                max_queue_delay_microseconds: 100
-            }
-            """
-
-        yaml_str = ("""
-            run_config_search_max_instance_count: 4
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = [
-            {'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': { 'max_queue_delay_microseconds': '100'}},
-            {'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': { 'max_queue_delay_microseconds': '100'}},
-            {'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': { 'max_queue_delay_microseconds': '100'}},
-            {'max_batch_size': 8, 'instance_group': [{'count': 4, 'kind': 'KIND_GPU'}],'dynamic_batching': { 'max_queue_delay_microseconds': '100'}}
-        ]
-        # yapf: enable
-
-        self._run_and_test_model_config_generator(yaml_str, expected_configs, protobuf)
-
-    def test_apply_value_to_dict(self):
-        """
-        Test different combinations of input and existing value types for apply_value_to_dict()
-        """
-        # Both input and existing are scalar value in a dict
-        existing_dict = {"a": 1, "b": 2}
-        expected_dict = {"a": 3, "b": 2}
-        BaseModelConfigGenerator._apply_value_to_dict("a", 3, existing_dict)
-        self.assertEqual(existing_dict, expected_dict)
-
-        # Input is scalar, existing is dict
-        existing_dict = {"a": 1, "b": {"c": 5, "d": 6}}
-        expected_dict = {"a": 1, "b": 2}
-        BaseModelConfigGenerator._apply_value_to_dict("b", 2, existing_dict)
-        self.assertEqual(existing_dict, expected_dict)
-
-        # Input is dict, existing is scalar
-        existing_dict = {"a": 1, "b": 3}
-        expected_dict = {"a": 1, "b": {"c": 7, "d": 8}}
-        BaseModelConfigGenerator._apply_value_to_dict(
-            "b", {"c": 7, "d": 8}, existing_dict
-        )
-        self.assertEqual(existing_dict, expected_dict)
-
-        # Input and dict are both dicts
-        existing_dict = {"a": 1, "b": {"c": {"e": 9, "f": 10}, "d": 6}}
-        expected_dict = {"a": 1, "b": {"c": {"e": 11, "f": 10}, "d": 6}}
-        BaseModelConfigGenerator._apply_value_to_dict(
-            "b", {"c": {"e": 11}}, existing_dict
-        )
-        self.assertEqual(existing_dict, expected_dict)
-
-    def test_early_exit_off_automatic_asserts(self):
-        """
-        Test that passing early_exit=False for automatic search raises an assert
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                - my-model
-            """)
-
-        expected_configs = []
-        # yapf: enable
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            self._run_and_test_model_config_generator(
-                yaml_str, expected_configs, early_exit_enable=False
-            )
-
-    def test_early_exit_on_automatic(self):
-        """
-        Test that automatic mode will early exit max_batch_size when throughput plateaus
-        """
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                - my-model
-            run_config_search_max_instance_count: 3
-            run_config_search_max_model_batch_size: 8
-            """)
-
-        expected_configs = [
-            {'max_batch_size': 1, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 1, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 1, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-            {'max_batch_size': 8, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}],'dynamic_batching': {}},
-        ]
-
-        with patch.object(TestModelConfigGenerator,
-                          "_get_next_fake_throughput") as mock_method:
-            mock_method.side_effect = [
-                1, 2, 4, 8,  # 1 instance
-                1, 1,        # 2 instances. Yes backoff
-                1, 2, 4, 8   # 3 instances
-            ]
-            self._run_and_test_model_config_generator(yaml_str,
-                                                      expected_configs,
-                                                      early_exit_enable=True)
-        # yapf: enable
-
-    def test_early_exit_off_manual(self):
-        """
-        Test that manual mode will not early exit despite throughput plateauing despite because early_exit_enable=False
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                test_model:
-                    model_config_parameters:
-                        max_batch_size: [1,2,4,8]
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: [1,2]
-            """)
-
-        expected_configs = [
-            {'max_batch_size': 1, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 8, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 1, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 4, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 8, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-        ]
-        # yapf: enable
-
-        with patch.object(
-            TestModelConfigGenerator, "_get_next_fake_throughput"
-        ) as mock_method:
-            mock_method.return_value = 1
-            self._run_and_test_model_config_generator(
-                yaml_str, expected_configs, early_exit_enable=False
-            )
-
-    def test_early_exit_on_manual(self):
-        """
-        Test that manual mode will early exit when throughput plateaus when early_exit_enable=True
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                test_model:
-                    model_config_parameters:
-                        max_batch_size: [1,2,3,4]
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: [1,2,3]
-            """)
-
-        expected_configs = [
-            {'max_batch_size': 1, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 2, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 3, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 4, 'instance_group': [{'count': 1, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 1, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 2, 'instance_group': [{'count': 2, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 1, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 2, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 3, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-            {'max_batch_size': 4, 'instance_group': [{'count': 3, 'kind': 'KIND_GPU'}]},
-        ]
-
-        with patch.object(TestModelConfigGenerator,
-                          "_get_next_fake_throughput") as mock_method:
-            mock_method.side_effect = [
-                1, 2, 4, 8,  # 1 instance
-                1, 1,        # 2 instances. Yes backoff
-                1, 2, 4, 8   # 3 instances
-            ]
-            self._run_and_test_model_config_generator(yaml_str,
-                                                      expected_configs,
-                                                      early_exit_enable=True)
-        # yapf: enable
-
-    def test_extract_model_name_from_variant_name(self):
-        input_output_pairs = {}
-        input_output_pairs[
-            "onnx_int32_int32_int32_config_default"
-        ] = "onnx_int32_int32_int32"
-        input_output_pairs["onnx_int32_int32_int32_config_2"] = "onnx_int32_int32_int32"
-        input_output_pairs["onnx_int32_int32_int32"] = "onnx_int32_int32_int32"
-
-        for variant_name, expected_model_name in input_output_pairs.items():
-            model_name = BaseModelConfigGenerator.extract_model_name_from_variant_name(
-                variant_name
-            )
-            self.assertEqual(model_name, expected_model_name)
-
-    def _run_and_test_model_config_generator(
-        self,
-        yaml_str,
-        expected_configs,
-        protobuf="max_batch_size: 8",
-        default_only=False,
-        early_exit_enable=True,
-    ):
-        """
-        Main function that creates a config from the yaml_str, runs it through
-        ModelConfigGenerator, and compares the resulting model_configs vs the expected_configs
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-
-        # Use mock model config or else TritonModelAnalyzerException will be thrown as it tries to read from disk
-        self.mock_model_config = MockModelConfig(protobuf)
-        self.mock_model_config.start()
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        profile_models = []
-        for model in config.profile_models:
-            profile_models.append(
-                ModelProfileSpec(model, config, MagicMock(), MagicMock())
-            )
-
-        # Fake out a client that can return a 'model_config' dict with
-        # a valid name (only used by remote mode)
-        #
-        fake_client = MagicMock()
-        fake_client.get_model_config = lambda name, retry_count: {"name": name}
-
-        try:
-            mcg = ModelConfigGeneratorFactory.create_model_config_generator(
-                config,
-                MagicMock(),
-                profile_models[0],
-                fake_client,
-                ModelVariantNameManager(),
-                default_only=default_only,
-                early_exit_enable=early_exit_enable,
-            )
-
-            model_configs = []
-            for model_config_variant in mcg.get_configs():
-                mcg.set_last_results(self._get_next_fake_results())
-                model_config_dict = model_config_variant.model_config.get_config()
-                model_configs.append(model_config_dict)
-
-            self.assertEqual(len(expected_configs), len(model_configs))
-
-            # Rip out the model name (so the order doesn't have to exactly match),
-            # but verify that it exists and is not none
-            #
-            for config in model_configs:
-                name = config.pop("name", None)
-                self.assertIsNotNone(name)
-
-            # Confirm the configs match
-            #
-            for config in expected_configs:
-                self.assertIn(config, model_configs)
-
-        finally:
-            self.mock_model_config.stop()
-
-    def _get_next_fake_results(self):
-        throughput = self._get_next_fake_throughput()
-
-        measurement = construct_run_config_measurement(
-            model_name=MagicMock(),
-            model_config_names=["test_model_config_name"],
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values=MagicMock(),
-            non_gpu_metric_values=[{"perf_throughput": throughput}],
-        )
-
-        return [measurement]
-
-    def _get_next_fake_throughput(self):
-        self._fake_throughput += 1
-        return self._fake_throughput
-
-    def setUp(self):
-        # Mock path validation
-        self.mock_os = MockOSMethods(
-            mock_paths=["model_analyzer.config.input.config_utils"]
-        )
-        self.mock_os.start()
-
-    def tearDown(self):
-        self.mock_os.stop()
-        patch.stopall()
-        ModelConfig._default_config_dict = {}
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_config_measurement.py b/src/c++/perf_analyzer/tests/test_model_config_measurement.py
deleted file mode 100755
index e760f1efa..000000000
--- a/src/c++/perf_analyzer/tests/test_model_config_measurement.py
+++ /dev/null
@@ -1,242 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import json
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.result.model_config_measurement import ModelConfigMeasurement
-from tests.common.test_utils import convert_non_gpu_metrics_to_data, default_encode
-
-from .common import test_result_collector as trc
-
-
-class TestModelConfigMeasurement(trc.TestResultCollector):
-    def setUp(self):
-        self.model_config_name = "modelA"
-        self.model_specific_pa_params = {"batch_size": 1, "concurrency-range": 1}
-
-        self.non_gpu_metric_values = {
-            "perf_throughput": 1000,
-            "perf_latency_p99": 20,
-            "cpu_used_ram": 1000,
-        }
-
-        self.mcmA = self._construct_model_config_measurement(
-            self.model_config_name,
-            self.model_specific_pa_params,
-            self.non_gpu_metric_values,
-        )
-
-        mcmB_non_gpu_metric_values = {
-            "perf_throughput": 2000,
-            "perf_latency_p99": 40,
-            "cpu_used_ram": 1000,
-        }
-
-        self.mcmB = self._construct_model_config_measurement(
-            "modelB", self.model_specific_pa_params, mcmB_non_gpu_metric_values
-        )
-
-        self.mcmC = self._construct_model_config_measurement(
-            "modelC", self.model_specific_pa_params, {}
-        )
-
-        self.mcmD = self._construct_model_config_measurement(
-            "modelD", self.model_specific_pa_params, {}
-        )
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_init(self):
-        """
-        Test that values are properly initialized
-        """
-        self.assertEqual(self.mcmA.model_config_name(), self.model_config_name)
-        self.assertEqual(
-            self.mcmA.model_specific_pa_params(), self.model_specific_pa_params
-        )
-        self.assertEqual(
-            self.mcmA.non_gpu_data(),
-            convert_non_gpu_metrics_to_data(self.non_gpu_metric_values),
-        )
-
-    def test_get_metric_found(self):
-        """
-        Test that non-gpu metrics can be correctly returned
-        """
-        non_gpu_data = convert_non_gpu_metrics_to_data(self.non_gpu_metric_values)
-
-        self.assertEqual(self.mcmA.get_metric("perf_throughput"), non_gpu_data[0])
-        self.assertEqual(self.mcmA.get_metric("perf_latency_p99"), non_gpu_data[1])
-        self.assertEqual(self.mcmA.get_metric("cpu_used_ram"), non_gpu_data[2])
-
-    def test_get_metric_not_found(self):
-        """
-        Test that an incorrect metric search returns None
-        """
-        self.assertEqual(self.mcmA.get_metric("XXXXX"), None)
-
-    def test_get_metric_value_found(self):
-        """
-        Test that non-gpu metric values can be correctly returned
-        """
-        self.assertEqual(
-            self.mcmA.get_metric_value("perf_throughput"),
-            self.non_gpu_metric_values["perf_throughput"],
-        )
-        self.assertEqual(
-            self.mcmA.get_metric_value("perf_latency_p99"),
-            self.non_gpu_metric_values["perf_latency_p99"],
-        )
-        self.assertEqual(
-            self.mcmA.get_metric_value("cpu_used_ram"),
-            self.non_gpu_metric_values["cpu_used_ram"],
-        )
-
-    def test_get_metric_value_not_found(self):
-        """
-        Test that an incorrect metric value search returns the correct value
-        """
-        self.assertEqual(self.mcmA.get_metric_value("XXXXX"), 0)
-        self.assertEqual(self.mcmA.get_metric_value("XXXXX", 100), 100)
-
-    def test_set_metric_weighting(self):
-        # Default
-        self.assertEqual({"perf_throughput": 1}, self.mcmA._metric_weights)
-
-        self.mcmA.set_metric_weighting({"perf_throughput": 2, "perf_latency_p99": 3})
-        expected_mw = {"perf_throughput": 2 / 5, "perf_latency_p99": 3 / 5}
-        self.assertEqual(expected_mw, self.mcmA._metric_weights)
-
-    def test_calculate_weighted_percentage_gain(self):
-        """
-        Test that weighted percentages are returned correctly
-        """
-        self.mcmA.set_metric_weighting({"perf_throughput": 1})
-        self.mcmB.set_metric_weighting({"perf_throughput": 1})
-
-        # throuhput: mcmA: 1000, mcmB: 2000
-        self.assertEqual(self.mcmA.calculate_weighted_percentage_gain(self.mcmB), -50)
-        self.assertEqual(self.mcmB.calculate_weighted_percentage_gain(self.mcmA), 100)
-
-        self.mcmA.set_metric_weighting({"perf_latency_p99": 1})
-        self.mcmB.set_metric_weighting({"perf_latency_p99": 1})
-
-        # latency: mcmA: 20, mcmB: 40
-        self.assertEqual(self.mcmA.calculate_weighted_percentage_gain(self.mcmB), 100)
-        self.assertEqual(self.mcmB.calculate_weighted_percentage_gain(self.mcmA), -50)
-
-        # This illustrates why we need to use score, not percentages to determine
-        # which model is better. In both cases we will (correctly) report that
-        # mcmA/B is 25% better than the other, even though they are equal
-        #
-        # mcmA has 50% worse throughput, but 100% better latency
-        # mcmB has 100% better latency, but 50% worse throughput
-        self.mcmA.set_metric_weighting({"perf_throughput": 1, "perf_latency_p99": 1})
-        self.mcmB.set_metric_weighting({"perf_throughput": 1, "perf_latency_p99": 1})
-        self.assertEqual(self.mcmA, self.mcmB)
-        self.assertEqual(self.mcmA.calculate_weighted_percentage_gain(self.mcmB), 25)
-        self.assertEqual(self.mcmB.calculate_weighted_percentage_gain(self.mcmA), 25)
-
-    def test_is_better_than(self):
-        """
-        Test that individual metric comparison works as expected
-        """
-        self.mcmA.set_metric_weighting({"perf_throughput": 1})
-
-        # throughput: 1000 is not better than 2000
-        self.assertFalse(self.mcmA.is_better_than(self.mcmB))
-        self.assertLess(self.mcmA, self.mcmB)
-
-        self.mcmA.set_metric_weighting({"perf_latency_p99": 1})
-
-        # latency: 20 is better than 40
-        self.assertTrue(self.mcmA.is_better_than(self.mcmB))
-        self.assertGreater(self.mcmA, self.mcmB)
-
-    def test_is_better_than_combo(self):
-        """
-        Test that combination metric comparison works as expected
-        """
-        # throuhput: 1000 vs. 2000 (worse), latency: 20 vs. 40 (better)
-        # with latency bias mcmA is better
-        self.mcmA.set_metric_weighting({"perf_throughput": 1, "perf_latency_p99": 3})
-
-        self.assertTrue(self.mcmA.is_better_than(self.mcmB))
-
-    def test_is_better_than_empty(self):
-        """
-        Test for correct return values when comparing for/against an empty set
-        """
-        self.mcmA.set_metric_weighting({"perf_throughput": 1})
-        self.mcmC.set_metric_weighting({"perf_throughput": 1})
-
-        self.assertTrue(self.mcmA.is_better_than(self.mcmC))
-        self.assertFalse(self.mcmC.is_better_than(self.mcmA))
-        self.assertEqual(self.mcmC, self.mcmD)
-
-    def test__eq__(self):
-        """
-        Test that individual metric equality works as expected
-        """
-        self.mcmA.set_metric_weighting({"cpu_used_ram": 10})
-
-        self.assertEqual(self.mcmA, self.mcmB)
-
-    def test__eq__combo(self):
-        """
-        Test that combination metric equality works as expected
-        """
-        # throuhput: 1000 vs. 2000 (worse), latency: 20 vs. 40 (better)
-        # with no bias they are equal
-        self.mcmA.set_metric_weighting({"perf_throughput": 1, "perf_latency_p99": 1})
-
-        self.assertEqual(self.mcmA, self.mcmB)
-
-    def test_from_dict(self):
-        """
-        Test to ensure class can be correctly restored from a dictionary
-        """
-        mcmA_json = json.dumps(self.mcmA, default=default_encode)
-
-        mcmA_from_dict = ModelConfigMeasurement.from_dict(json.loads(mcmA_json))
-
-        self.assertEqual(
-            mcmA_from_dict.model_config_name(), self.mcmA.model_config_name()
-        )
-        self.assertEqual(
-            mcmA_from_dict.model_specific_pa_params(),
-            self.mcmA.model_specific_pa_params(),
-        )
-        self.assertEqual(mcmA_from_dict.non_gpu_data(), self.mcmA.non_gpu_data())
-
-        # Catchall in case something new is added
-        self.assertEqual(mcmA_from_dict, self.mcmA)
-
-    def _construct_model_config_measurement(
-        self, model_config_name, model_specific_pa_params, non_gpu_metric_values
-    ):
-        non_gpu_data = convert_non_gpu_metrics_to_data(non_gpu_metric_values)
-
-        return ModelConfigMeasurement(
-            model_config_name, model_specific_pa_params, non_gpu_data
-        )
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_constraints.py b/src/c++/perf_analyzer/tests/test_model_constraints.py
deleted file mode 100755
index a265f3bdb..000000000
--- a/src/c++/perf_analyzer/tests/test_model_constraints.py
+++ /dev/null
@@ -1,119 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.result.model_constraints import ModelConstraints
-
-from .common import test_result_collector as trc
-
-
-class TestModelConstraints(trc.TestResultCollector):
-    def tearDown(self):
-        patch.stopall()
-
-    def test_to_dict(self):
-        """
-        Test that constraints object to dict
-        """
-        constraints_dict_1 = {
-            "perf_latency_p99": {"max": 100},
-            "gpu_used_memory": {"max": 20000},
-        }
-        constraints_dict_2 = {}
-
-        constraints_obj_1 = ModelConstraints(constraints=constraints_dict_1)
-        constraints_obj_2 = ModelConstraints(constraints=constraints_dict_2)
-
-        self.assertEqual(constraints_obj_1.to_dict(), constraints_dict_1)
-        self.assertEqual(constraints_obj_2.to_dict(), constraints_dict_2)
-
-    def test_has_metric(self):
-        """
-        Test that constraints object has a given key
-        """
-        constraints_dict_1 = {
-            "perf_latency_p99": {"max": 100},
-            "gpu_used_memory": {"max": 20000},
-        }
-        constraints_dict_2 = {}
-
-        constraints_obj_1 = ModelConstraints(constraints=constraints_dict_1)
-        constraints_obj_2 = ModelConstraints(constraints=constraints_dict_2)
-
-        self.assertTrue(constraints_obj_1.has_metric("perf_latency_p99"))
-        self.assertTrue(constraints_obj_1.has_metric("gpu_used_memory"))
-        self.assertFalse(constraints_obj_1.has_metric("perf_throughput"))
-        self.assertFalse(constraints_obj_1.has_metric(""))
-        self.assertFalse(constraints_obj_2.has_metric("perf_latency_p99"))
-        self.assertFalse(constraints_obj_2.has_metric("gpu_used_memory"))
-
-    def test_getitem(self):
-        """
-        Test that metrics are subscriptable from constraints object
-        """
-        constraints_dict = {
-            "perf_latency_p99": {"max": 100},
-            "gpu_used_memory": {"max": 20000},
-        }
-
-        constraints_obj = ModelConstraints(constraints=constraints_dict)
-
-        self.assertEqual(constraints_obj["perf_latency_p99"], {"max": 100})
-        self.assertEqual(constraints_obj["gpu_used_memory"], {"max": 20000})
-
-    def test_bool(self):
-        """
-        Test that constraints are empty
-        """
-        constraints_dict_1 = {
-            "perf_latency_p99": {"max": 100},
-            "gpu_used_memory": {"max": 20000},
-        }
-        constraints_dict_2 = {}
-
-        constraints_obj_1 = ModelConstraints(constraints=constraints_dict_1)
-        constraints_obj_2 = ModelConstraints(constraints=constraints_dict_2)
-
-        self.assertTrue(constraints_obj_1)
-        self.assertFalse(constraints_obj_2)
-
-    def test_eq(self):
-        """
-        Test that constraint objects are equal
-        """
-        constraints_dict_1 = {
-            "perf_latency_p99": {"max": 100},
-            "gpu_used_memory": {"max": 20000},
-        }
-        constraints_dict_2 = {
-            "perf_latency_p99": {"max": 100},
-            "gpu_used_memory": {"max": 20000},
-        }
-        constraints_dict_3 = {}
-
-        constraints_obj_1 = ModelConstraints(constraints=constraints_dict_1)
-        constraints_obj_2 = ModelConstraints(constraints=constraints_dict_2)
-        constraints_obj_3 = ModelConstraints(constraints=constraints_dict_3)
-
-        self.assertEqual(constraints_obj_1, constraints_obj_2)
-        self.assertNotEqual(constraints_obj_3, constraints_obj_2)
-        self.assertNotEqual(constraints_obj_1, constraints_obj_3)
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_manager.py b/src/c++/perf_analyzer/tests/test_model_manager.py
deleted file mode 100755
index 0370c6c77..000000000
--- a/src/c++/perf_analyzer/tests/test_model_manager.py
+++ /dev/null
@@ -1,1319 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.config.input.objects.config_model_profile_spec import (
-    ConfigModelProfileSpec,
-)
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.model_manager import ModelManager
-from model_analyzer.record.metrics_manager import MetricsManager
-from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
-from model_analyzer.triton.model.model_config import ModelConfig
-
-from .common import test_result_collector as trc
-from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
-from .mocks.mock_model_config import MockModelConfig
-from .mocks.mock_run_configs import MockRunConfigs
-
-
-class MetricsManagerSubclass(MetricsManager):
-    """
-    Overrides execute_model_run_config() to gather a list of MockRunConfigs that
-    contain the configured values of each would-be 'executed' run_config
-    """
-
-    def __init__(self, config, client, server, gpus, result_manager, state_manager):
-        super().__init__(config, client, server, gpus, result_manager, state_manager)
-        self._configs = MockRunConfigs()
-        self._perf_throughput = 1
-
-    def get_run_configs(self):
-        """Return the list of configs that would have been 'executed'"""
-        return self._configs
-
-    def execute_run_config(self, config):
-        self._configs.add_from_model_run_config(config.model_run_configs()[0])
-        return self._get_next_measurements()
-
-    def _get_next_measurements(self):
-        """Return fake measurements as if the run_configs had been executed"""
-
-        throughput_value = self._get_next_perf_throughput_value()
-        if throughput_value is None:
-            return None
-        else:
-            return construct_run_config_measurement(
-                model_name=MagicMock(),
-                model_config_names=["test_model_config_name"],
-                model_specific_pa_params=MagicMock(),
-                gpu_metric_values=MagicMock(),
-                non_gpu_metric_values=[{"perf_throughput": throughput_value}],
-            )
-
-    def _get_next_perf_throughput_value(self):
-        self._perf_throughput *= 2
-        return self._perf_throughput
-
-
-class TestModelManager(trc.TestResultCollector):
-    def tearDown(self):
-        patch.stopall()
-        ModelConfig._default_config_dict = {}
-
-    def __init__(self, methodname):
-        super().__init__(methodname)
-        self._args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-
-        self._model_config_protobuf = """
-            name: "test_model"
-            platform: "tensorflow_graphdef"
-            max_batch_size: 8
-            input [
-            {
-                name: "INPUT__0"
-                data_type: TYPE_FP32
-                dims: [16]
-            }
-            ]
-            output [
-            {
-                name: "OUTPUT__0"
-                data_type: TYPE_FP32
-                dims: [16]
-            }
-            ]
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 1
-            }
-            ]
-            """
-
-    def test_full_sweep(self):
-        """
-        Test a normal full sweep of options
-        """
-        expected_ranges = [
-            {
-                "instances": [1, 2, 3, 4, 5],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1],
-                "max_batch_size": [1, 2, 4, 8],
-                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 128
-            run_config_search_max_instance_count: 5
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_another_full_sweep(self):
-        """
-        Test another full sweep of options
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1, 2, 3, 4, 5, 6, 7],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8, 16, 32],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8, 16, 32],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 32
-            run_config_search_max_instance_count: 7
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_run_search_disable(self):
-        """
-        Test with run_config_search_disable=True
-
-        Expect 1 result that matches the default configuration because no manual
-        search options provided and automatic search disabled/ignored
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 32
-            run_config_search_max_instance_count: 7
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: True
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_manual_concurrency(self):
-        """
-        Test with manually specified concurrencies
-        """
-        expected_ranges = [
-            {
-                "instances": [1, 2, 3, 4, 5, 6, 7],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [5, 7],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [5, 7],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 32
-            run_config_search_max_instance_count: 7
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            concurrency: [5, 7]
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_remote_mode(self):
-        """
-        Test remote mode
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1, 2],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8, 16],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8, 16],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 16
-            run_config_search_max_instance_count: 2
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            triton_launch_mode: remote
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_manual_parameters(self):
-        """
-        Test with manually specified concurrencies and batch sizes
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1, 2, 3, 4, 5, 6, 7],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1, 2, 3],
-                "max_batch_size": [8],
-                "concurrency": [2, 10, 18, 26, 34, 42, 50, 58],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1, 2, 3],
-                "max_batch_size": [8],
-                "concurrency": [2, 10, 18, 26, 34, 42, 50, 58],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 512
-            run_config_search_max_instance_count: 7
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            concurrency:
-                start: 2
-                stop: 64
-                step: 8
-            batch_sizes: 1,2,3
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_no_early_exit_client_batch_size(self):
-        """
-        Test that we will NOT early exit client batch size by default
-        """
-        self._test_early_exit_client_batch_size_helper(early_exit=False)
-
-    def test_early_exit_client_batch_size(self):
-        """
-        Test that we will early exit client batch size if
-        --early-exit-enable is true
-        """
-        self._test_early_exit_client_batch_size_helper(early_exit=True)
-
-    def _test_early_exit_client_batch_size_helper(self, early_exit):
-        args = self._args.copy()
-
-        if early_exit:
-            args.append("--early-exit-enable")
-
-            expected_ranges = [
-                {
-                    "instances": [1],
-                    "kind": ["KIND_GPU"],
-                    "batching": [0],
-                    "batch_sizes": [1, 2, 3, 4],
-                    "max_batch_size": [8],
-                    "concurrency": [1, 2, 4, 8, 16],
-                },
-                {
-                    "instances": [1],
-                    "kind": ["KIND_CPU"],
-                    "batching": [None],
-                    "batch_sizes": [1, 2, 3, 4, 7],
-                    "max_batch_size": [8],
-                    "concurrency": [1, 2, 4, 8],
-                },
-            ]
-
-        else:
-            expected_ranges = [
-                {
-                    "instances": [1],
-                    "kind": ["KIND_GPU"],
-                    "batching": [0],
-                    "batch_sizes": [1, 2, 3, 4, 7],
-                    "max_batch_size": [8],
-                    "concurrency": [1, 2, 4, 8, 16],
-                },
-                {
-                    "instances": [1],
-                    "kind": ["KIND_CPU"],
-                    "batching": [None],
-                    "batch_sizes": [1, 2, 3, 4, 7],
-                    "max_batch_size": [8],
-                    "concurrency": [1, 2, 4, 8, 16],
-                },
-            ]
-
-        yaml_content = """
-            profile_models: test_model
-            run_config_search_max_instance_count: 1
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            concurrency: 16,1,2,4,8
-            batch_sizes: 3,7,1,4,2
-            """
-
-        with patch.object(
-            MetricsManagerSubclass, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            # yapf: disable
-            side_effect = [
-                # Default config, bs=1, concurrency 1,2,4,8
-                # Will early exit for concurrency
-                # "Best" result for bs early exit is 5
-                5, 5, 5, 5,
-
-                # Default config, bs=2, concurrency 1,2,4,8
-                # Will early exit for concurrency
-                # "Best" result for bs early exit is 4
-                4, 4, 4, 4,
-
-                # Default config, bs=3, concurrency 1,2,4,8
-                # Will early exit for concurrency
-                # "Best" result for bs early exit is 6
-                6, 6, 6, 6,
-
-                # Default config, bs=4, concurrency 1,2,4,8
-                # Will early exit for concurrency
-                # "Best" result for bs early exit is 5
-                # We will not early exit batch size
-                5, 5, 5, 5,
-
-                # Default config, bs=7, concurrency 1,2,4,8
-                # Will not early exit for concurrency
-                # "Best" result for bs early exit is 1
-                # We are done sweeping batch size
-                1, 1, 1, 1,
-
-
-                # 1 instance, bs=1, concurrency 1,2,4,8,16
-                # Will not early exit for concurrency
-                # "Best" result for bs early exit is 10
-                1, 1, 10, 1, 1,
-
-                # 1 instance, bs=2, concurrency 1,2,4,8,16
-                # Will not early exit for concurrency
-                # "Best" result for bs early exit is 9
-                1, 9, 1, 1, 2,
-
-                # 1 instance, bs=3, concurrency 1,2,4,8,16
-                # Will not early exit for concurrency
-                # "Best" result for bs early exit is 8
-                1, 1, 1, 8, 3,
-
-                # 1 instance, bs=4, concurrency 1,2,4,8,16
-                # Will not early exit for concurrency
-                # "Best" result for bs early exit is 7
-                # Will early exit batch size now
-                1, 1, 7, 1, 4
-            ]
-            # Add a bunch of extra results for the no-early-exit case
-            side_effect.extend([1]*100)
-            mock_method.side_effect = side_effect
-            # yapf: enable
-
-            self._test_model_manager(yaml_content, expected_ranges, args=args)
-
-    def test_model_config_parameters(self):
-        """
-        Test with manually specified model config parameters
-
-        In this case we don't automatically search instances or dynamic_batching
-        since model config parameters are specified.
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batch_sizes": [1],
-                "max_batch_size": [1, 3, 7, 8, 13],
-                "concurrency": [1, 2, 4, 8],
-            }
-        ]
-
-        yaml_str = """
-            run_config_search_max_concurrency: 8
-            run_config_search_max_instance_count: 16
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            profile_models:
-                test_model:
-                    model_config_parameters:
-                        max_batch_size: [1,3,7,13]
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_default_config_always_run_no_dynamic_batching_off(self):
-        """
-        Test that the default config is run even when manual search excludes that case
-        In this case, default config is (1 instance, max_batch_size=8, dynamic batching off)
-        We should have a case of dynamic_batching off even though manual search only has it on
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [0],
-                "max_queue_delay": ["200", "300"],
-                "batch_sizes": [1],
-                "max_batch_size": [1, 2, 4, 8, 16],
-                "concurrency": [1, 2, 4, 8],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8],
-            },
-        ]
-
-        yaml_str = """
-            run_config_search_max_concurrency: 8
-            run_config_search_max_instance_count: 16
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            profile_models:
-                test_model:
-                    model_config_parameters:
-                        max_batch_size: [1,2,4,8,16]
-                        dynamic_batching:
-                            max_queue_delay_microseconds: [200, 300]
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_default_config_always_run_wrong_instances(self):
-        """
-        Test that the default config is run even when manual search excludes that case
-        In this case, default config is (2 instances, max_batch_size=8, dynamic batching off)
-        We should have a 2-instance case even though manual search only has 1-instance
-        """
-
-        self._model_config_protobuf = """
-            name: "test_model"
-            platform: "tensorflow_graphdef"
-            max_batch_size: 8
-            input [
-            {
-                name: "INPUT__0"
-                data_type: TYPE_FP32
-                dims: [16]
-            }
-            ]
-            output [
-            {
-                name: "OUTPUT__0"
-                data_type: TYPE_FP32
-                dims: [16]
-            }
-            ]
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 2
-            }
-            ]
-            """
-
-        expected_ranges = [
-            {
-                "instances": [1],
-                "kind": ["KIND_GPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4],
-            },
-            {
-                "instances": [2],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4],
-            },
-        ]
-
-        yaml_str = """
-            run_config_search_max_concurrency: 4
-            run_config_search_max_instance_count: 16
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            profile_models:
-                test_model:
-                    model_config_parameters:
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: 1
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_default_config_always_run_cpu_vs_gpu(self):
-        """
-        If the default configuration had KIND_CPU, make sure it is run (even if everything
-        else is the same)
-        """
-
-        self._model_config_protobuf = """
-            name: "test_model"
-            platform: "tensorflow_graphdef"
-            max_batch_size: 8
-            input [
-            {
-                name: "INPUT__0"
-                data_type: TYPE_FP32
-                dims: [16]
-            }
-            ]
-            output [
-            {
-                name: "OUTPUT__0"
-                data_type: TYPE_FP32
-                dims: [16]
-            }
-            ]
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 1
-            }
-            ]
-            """
-
-        expected_ranges = [
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_GPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4],
-            },
-        ]
-
-        yaml_str = """
-            run_config_search_max_concurrency: 4
-            run_config_search_max_instance_count: 16
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            profile_models:
-                test_model:
-                    model_config_parameters:
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: 1
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_default_config_always_run_automatic_search(self):
-        """
-        Test that the default config is run even when automatic search excludes that case
-        In this case, default config is (4 instance, CPU, max_batch_size=8, dynamic batching off)
-        We should have this 4 instance case though run_config_search_max_instance_count=1
-        """
-
-        self._model_config_protobuf = """
-            name: "test_model"
-            platform: "tensorflow_graphdef"
-            max_batch_size: 8
-            input [
-            {
-                name: "INPUT__0"
-                data_type: TYPE_FP32
-                dims: [16]
-            }
-            ]
-            output [
-            {
-                name: "OUTPUT__0"
-                data_type: TYPE_FP32
-                dims: [16]
-            }
-            ]
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 4
-            }
-            ]
-            """
-
-        expected_ranges = [
-            {
-                "instances": [1],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4],
-            },
-            {
-                "instances": [4],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4],
-            },
-        ]
-
-        yaml_str = """
-            run_config_search_max_concurrency: 4
-            run_config_search_max_instance_count: 1
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            profile_models: test_model
-            """
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_throughput_early_exit_minimum_runs(self):
-        """
-        Test that there is an early back off when sweeping concurrency
-
-        The behavior is that MA will try at least 4 concurrencies. If
-        at that point none of the last 3 attempts have had satisfactory
-        gain, it will stop
-
-        This test hardcodes the 'throughput' to 1, so for all model
-        configs the gain will be invalid and it will only try 4
-        concurrencies of (1,2,4,8) despite max_concurrency=128
-
-        This test also has multiple batch sizes to make sure that hitting
-        an early exit on one doesn't cause the next one to be skipped
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1, 2],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1, 2],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1, 2],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 128
-            run_config_search_max_instance_count: 2
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            batch_sizes: 1,2
-            """
-
-        with patch.object(
-            MetricsManagerSubclass, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.return_value = 1
-            self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_no_early_exit_if_not_auto_search(self):
-        """
-        Test that there is NOT an early back off when sweeping concurrency if not in auto sweep mode
-
-        This test hardcodes the 'throughput' to 1, so for all model
-        configs the gain will be invalid. However, it should still sweep
-        to concurrency of 128 due to the fact that it was manually specified instead
-        of an auto-search
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1, 2],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1, 2],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1, 2],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 128
-            run_config_search_max_instance_count: 2
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            batch_sizes: 1,2
-            concurrency: 1,2,4,8,16,32,64,128
-            """
-
-        with patch.object(
-            MetricsManagerSubclass, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.return_value = 1
-            self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_throughput_early_exit(self):
-        """
-        Test that there is an early back off when sweeping concurrency
-
-        The behavior is that MA stop if it had 4 concurrencies in a row
-        without any valid gain amongst any of them
-
-        This test sets the 'throughput' to [1,2,4,8,16,16,16,16], which
-        will cause an early exit after trying the 8th concurrency (128)
-        instead of searching all the way to 2048
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8, 16, 32, 64, 128],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 2048
-            run_config_search_max_instance_count: 1
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            """
-
-        with patch.object(
-            MetricsManagerSubclass, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.side_effect = [
-                1,
-                2,
-                4,
-                8,
-                16,
-                16,
-                16,
-                16,
-                1,
-                2,
-                4,
-                8,
-                16,
-                16,
-                16,
-                16,
-            ]
-            self._test_model_manager(yaml_str, expected_ranges)
-
-    @patch("model_analyzer.model_manager.INVALID_MEASUREMENT_THRESHOLD", 999)
-    def test_bad_result_early_PA_exit(self):
-        """
-        Test that there is an early back off for bad result (out of memory)
-
-        If no measurements are returned in an attempt, no further concurrencies
-        should be tried.
-
-        This test hardcodes the measurements to be empty (bad result), so for all
-        model configs it will only try 1 concurrency despite max_concurrency=128
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1, 2],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 128
-            run_config_search_max_instance_count: 2
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            """
-
-        with patch.object(
-            MetricsManagerSubclass, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.return_value = None
-            self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_report_failure_no_measurements(self):
-        """
-        Test that MA takes an exception if we detect no measurements returned from
-        PA at the start of profile
-        """
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 128
-            run_config_search_max_instance_count: 2
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 8
-            run_config_search_disable: False
-            """
-
-        with patch.object(
-            MetricsManagerSubclass, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.return_value = None
-            with self.assertRaises(TritonModelAnalyzerException):
-                self._test_model_manager(yaml_str, None)
-
-    def test_lower_throughput_early_batch_size_exit(self):
-        """
-        Test that there is an early back off for throughput decreasing
-        when sweeping max_batch_size
-
-        If a list of measurements is provided with a lower max throughput than the previous
-        list of measurements, then we should early exit that max_batch_size sweep
-
-        The test is set up such that at instance count 1, going from max_batch_size of 16 to 32
-        does not increase the throughput, and thus we should not continue stepping max_batch_size.
-        The same is true for instance count 2, going from max_batch_size of 32 to 64
-        """
-
-        expected_ranges = [
-            # Instance count of 1 will stop after max_batch_size=32
-            {
-                "instances": [1],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1],
-                "max_batch_size": [8, 16, 32],
-                "concurrency": [1, 2, 4],
-            },
-            # Instance count of 2 will stop after max_batch_size=64
-            {
-                "instances": [2],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1],
-                "max_batch_size": [8, 16, 32, 64],
-                "concurrency": [1, 2, 4],
-            },
-            # Default config
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 4
-            run_config_search_max_instance_count: 2
-            run_config_search_min_model_batch_size: 8
-            run_config_search_disable: False
-            """
-
-        with patch.object(
-            MetricsManagerSubclass, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            # yapf: disable
-            mock_method.side_effect = [
-                1, 2, 4,     # Default config, concurrency 1,2,4
-                1, 2, 4,     # 1 Instance, Batch size 8, concurrency 1,2,4
-                2, 4, 8,     # 1 Instance, Batch size 16, concurrency 1,2,4
-                2, 4, 8,     # 1 Instance, Batch size 32, concurrency 1,2,4
-                1, 2, 4,     # 1 Instance, Batch size 8, concurrency 1,2,4
-                8, 4, 2,     # 1 Instance, Batch size 16, concurrency 1,2,4
-                4, 8, 16,    # 1 Instance, Batch size 32, concurrency 1,2,4
-                4, 8, 16     # 1 Instance, Batch size 64, concurrency 1,2,4
-            ]
-            # yapf: enable
-
-            mock_method.return_value = None
-            self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_no_max_batch_size_sweep_if_protobuf_0(self):
-        """
-        Test that if the max_batch_size is 0 in the original model config,
-        then do not sweep max batch size or dynamic batching
-        """
-
-        self._model_config_protobuf = """
-            name: "test_model"
-            platform: "tensorflow_graphdef"
-            max_batch_size: 0
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 1
-            }
-            ]
-            """
-
-        expected_ranges = [
-            {
-                "instances": [1],
-                "kind": ["KIND_GPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "concurrency": [1, 2, 4],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1],
-                "concurrency": [1, 2, 4],
-            },
-        ]
-
-        yaml_str = """
-            run_config_search_max_concurrency: 4
-            run_config_search_max_instance_count: 1
-            run_config_search_disable: False
-            profile_models: test_model
-            """
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    def test_client_batch_size_never_above_max_batch_size(self):
-        """
-        When we are sweeping through values of PA batch size and model max_batch_size,
-        it should never be the case that PA batch size > model max_batch_size
-        """
-
-        expected_ranges = [
-            {
-                "instances": [1],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1, 2, 4, 8],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1, 2, 4, 8, 16],
-                "max_batch_size": [16],
-                "concurrency": [1, 2, 4, 8],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_GPU"],
-                "batching": [0],
-                "batch_sizes": [1, 2, 4, 8, 16, 32],
-                "max_batch_size": [32],
-                "concurrency": [1, 2, 4, 8],
-            },
-            {
-                "instances": [1],
-                "kind": ["KIND_CPU"],
-                "batching": [None],
-                "batch_sizes": [1, 2, 4, 8],
-                "max_batch_size": [8],
-                "concurrency": [1, 2, 4, 8],
-            },
-        ]
-
-        yaml_str = """
-            profile_models: test_model
-            run_config_search_max_concurrency: 8
-            run_config_search_max_instance_count: 1
-            run_config_search_min_model_batch_size: 8
-            run_config_search_max_model_batch_size: 32
-            run_config_search_disable: False
-            batch_sizes: 1,2,4,8,16,32,64
-            """
-
-        self._test_model_manager(yaml_str, expected_ranges)
-
-    @patch(
-        "model_analyzer.triton.model.model_config.ModelConfig.is_ensemble",
-        return_value=True,
-    )
-    def test_ensemble_illegal_checks(self, *args):
-        """
-        Test that RCS mode isn't set to brute and that multiple models are
-        not provided when profiling an ensemble model
-        """
-        yaml_str = """
-                  profile_models: ensemble_model, test_model
-                  """
-
-        args = self._args.copy()
-        args.append("--run-config-search-mode")
-        args.append("brute")
-
-        self.mock_model_config = MockModelConfig(self._model_config_protobuf)
-        self.mock_model_config.start()
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        state_manager = AnalyzerStateManager(config, MagicMock())
-        metrics_manager = MetricsManagerSubclass(
-            config, MagicMock(), MagicMock(), MagicMock(), MagicMock(), state_manager
-        )
-        model_manager = ModelManager(
-            config,
-            MagicMock(),
-            MagicMock(),
-            MagicMock(),
-            metrics_manager,
-            MagicMock(),
-            state_manager,
-            MagicMock(),
-        )
-
-        # Multiple model check
-        models = [
-            ConfigModelProfileSpec("ensemble_model"),
-            ConfigModelProfileSpec("test_model"),
-        ]
-        with self.assertRaises(TritonModelAnalyzerException):
-            model_manager._check_for_ensemble_model_incompatibility(models)
-
-        # RunConfigSearch check
-        models = [
-            ConfigModelProfileSpec("ensemble_model"),
-        ]
-        with self.assertRaises(TritonModelAnalyzerException):
-            model_manager._check_for_ensemble_model_incompatibility(models)
-        self.mock_model_config.stop()
-
-    @patch(
-        "model_analyzer.triton.model.model_config.ModelConfig.is_ensemble",
-        return_value=True,
-    )
-    def test_ensemble_makes_quick_default(self, *args):
-        """
-        Test that the default RCS mode is switched from brute to quick
-        when profiling an ensemble model
-        """
-        yaml_str = """
-                  profile_models: ensemble_model
-                  """
-
-        args = self._args.copy()
-
-        self.mock_model_config = MockModelConfig(self._model_config_protobuf)
-        self.mock_model_config.start()
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        state_manager = AnalyzerStateManager(config, MagicMock())
-        metrics_manager = MetricsManagerSubclass(
-            config, MagicMock(), MagicMock(), MagicMock(), MagicMock(), state_manager
-        )
-        model_manager = ModelManager(
-            config,
-            MagicMock(),
-            MagicMock(),
-            MagicMock(),
-            metrics_manager,
-            MagicMock(),
-            state_manager,
-            MagicMock(),
-        )
-
-        models = [
-            ConfigModelProfileSpec("ensemble_model"),
-        ]
-        model_manager._check_for_ensemble_model_incompatibility(models)
-
-        self.assertEqual(config.run_config_search_mode, "quick")
-
-    @patch(
-        "model_analyzer.triton.model.model_config.ModelConfig.is_ensemble",
-        return_value=False,
-    )
-    def test_cpu_only_composing_models_error(self, *args):
-        """
-        Test that --cpu-only-composing-models errors when
-        set for non-ensemble/BLS models
-        """
-        yaml_str = """
-                  profile_models: test_model
-                  """
-
-        args = self._args.copy()
-        args.append("--cpu-only-composing-models")
-        args.append("composing_modelA,composing_modelB")
-
-        self.mock_model_config = MockModelConfig(self._model_config_protobuf)
-        self.mock_model_config.start()
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        state_manager = AnalyzerStateManager(config, MagicMock())
-        metrics_manager = MetricsManagerSubclass(
-            config, MagicMock(), MagicMock(), MagicMock(), MagicMock(), state_manager
-        )
-        model_manager = ModelManager(
-            config,
-            MagicMock(),
-            MagicMock(),
-            MagicMock(),
-            metrics_manager,
-            MagicMock(),
-            state_manager,
-            MagicMock(),
-        )
-
-        # RunConfigSearch check
-        models = [
-            ConfigModelProfileSpec("test_model"),
-        ]
-        with self.assertRaises(TritonModelAnalyzerException):
-            model_manager._check_for_ensemble_model_incompatibility(models)
-        self.mock_model_config.stop()
-
-    def _test_model_manager(self, yaml_content, expected_ranges, args=None):
-        """
-        Test helper function that passes the given yaml_str into
-        model_manager, runs the model, and confirms the result is as expected
-        based on a full cartesian product of the lists in the input list of
-        dicts expected_ranges
-        """
-
-        if args is None:
-            args = self._args
-
-        # Use mock model config or else TritonModelAnalyzerException will be thrown as it tries to read from disk
-        self.mock_model_config = MockModelConfig(self._model_config_protobuf)
-        self.mock_model_config.start()
-        config = evaluate_mock_config(args, yaml_content, subcommand="profile")
-
-        state_manager = AnalyzerStateManager(config, MagicMock())
-        metrics_manager = MetricsManagerSubclass(
-            config, MagicMock(), MagicMock(), MagicMock(), MagicMock(), state_manager
-        )
-        model_manager = ModelManager(
-            config,
-            MagicMock(),
-            MagicMock(),
-            MagicMock(),
-            metrics_manager,
-            MagicMock(),
-            state_manager,
-            MagicMock(),
-        )
-
-        model_manager.run_models([config.profile_models[0]])
-        self.mock_model_config.stop()
-
-        self._check_results(model_manager, expected_ranges)
-
-    def _check_results(self, model_manager, expected_ranges):
-        """
-        Create a set of expected and actual run configs and confirm they are equal
-        """
-        run_configs = model_manager._metrics_manager.get_run_configs()
-        expected_configs = MockRunConfigs()
-        expected_configs.populate_from_ranges(expected_ranges)
-
-        self.assertEqual(
-            run_configs.get_configs_set(), expected_configs.get_configs_set()
-        )
-        self.assertEqual(
-            run_configs.get_num_configs(), expected_configs.get_num_configs()
-        )
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_model_variant_name_manager.py b/src/c++/perf_analyzer/tests/test_model_variant_name_manager.py
deleted file mode 100755
index b21e57404..000000000
--- a/src/c++/perf_analyzer/tests/test_model_variant_name_manager.py
+++ /dev/null
@@ -1,266 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.config.generate.model_variant_name_manager import (
-    ModelVariantNameManager,
-)
-from model_analyzer.constants import DEFAULT_CONFIG_PARAMS
-from model_analyzer.triton.model.model_config import ModelConfig
-from model_analyzer.triton.model.model_config_variant import ModelConfigVariant
-from tests.common.test_utils import default_encode
-
-from .common import test_result_collector as trc
-
-
-class TestModelVariantNameManager(trc.TestResultCollector):
-    def setUp(self):
-        self._mvnm = ModelVariantNameManager()
-        self._non_default_param_combo = {"foo": 1}
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_default(self):
-        """
-        Check that default config is returned
-        """
-        name = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 1}, DEFAULT_CONFIG_PARAMS
-        )
-
-        self.assertEqual(name, (False, "modelA_config_default"))
-
-    def test_basic(self):
-        """
-        If multiple unique model configs are passed in, the name will keep
-        incrementing
-        """
-        a0 = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 1}, self._non_default_param_combo
-        )
-        a1 = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 2}, self._non_default_param_combo
-        )
-        a2 = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 4}, self._non_default_param_combo
-        )
-
-        self.assertEqual(a0, (False, "modelA_config_0"))
-        self.assertEqual(a1, (False, "modelA_config_1"))
-        self.assertEqual(a2, (False, "modelA_config_2"))
-
-    def test_multiple_models(self):
-        """
-        The two models should have no impact on each other's naming or counts
-        """
-
-        a0 = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 1}, self._non_default_param_combo
-        )
-        b0 = self._mvnm.get_model_variant_name(
-            "modelB", {"A": 1}, self._non_default_param_combo
-        )
-        a1 = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 2}, self._non_default_param_combo
-        )
-        b1 = self._mvnm.get_model_variant_name(
-            "modelB", {"A": 2}, self._non_default_param_combo
-        )
-
-        self.assertEqual(a0, (False, "modelA_config_0"))
-        self.assertEqual(a1, (False, "modelA_config_1"))
-        self.assertEqual(b0, (False, "modelB_config_0"))
-        self.assertEqual(b1, (False, "modelB_config_1"))
-
-    def test_repeat(self):
-        """
-        Calling with the same model name/config/combo multiple times
-        should result in the same config name being returned
-        """
-
-        a0 = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 1}, self._non_default_param_combo
-        )
-        a1 = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 1}, self._non_default_param_combo
-        )
-
-        self.assertEqual(a0, (False, "modelA_config_0"))
-        self.assertEqual(a1, (True, "modelA_config_0"))
-
-    def test_nested_dicts_matching(self):
-        """
-        Test matching with a model config consisting of matching nested dicts
-        """
-        model_config_A_0 = {"A": {"B": {"C": 1, "D": 2}}}
-        model_config_A_1 = {"A": {"B": {"D": 2, "C": 1}}}
-
-        a0 = self._mvnm.get_model_variant_name(
-            "modelA", model_config_A_0, self._non_default_param_combo
-        )
-
-        a1 = self._mvnm.get_model_variant_name(
-            "modelA", model_config_A_1, self._non_default_param_combo
-        )
-
-        self.assertEqual(a0, (False, "modelA_config_0"))
-        self.assertEqual(a1, (True, "modelA_config_0"))
-
-    def test_nested_dicts_different(self):
-        """
-        Test matching with a model config consisting of different nested dicts
-        """
-        model_config_A_0 = {"A": {"B": {"C": 1, "D": 2}}}
-        model_config_A_1 = {"A": {"B": {"D": 1, "C": 2}}}
-
-        a0 = self._mvnm.get_model_variant_name(
-            "modelA", model_config_A_0, self._non_default_param_combo
-        )
-
-        a1 = self._mvnm.get_model_variant_name(
-            "modelA", model_config_A_1, self._non_default_param_combo
-        )
-
-        self.assertEqual(a0, (False, "modelA_config_0"))
-        self.assertEqual(a1, (False, "modelA_config_1"))
-
-    def test_ensemble_default(self):
-        """
-        Test that a default ensemble config is returned
-        """
-        sub_configA = ModelConfigVariant(ModelConfig({}), "modelA_config_default")
-        sub_configB = ModelConfigVariant(ModelConfig({}), "modelB_config_default")
-
-        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
-            [sub_configA, sub_configB]
-        )
-
-        name = self._mvnm.get_ensemble_model_variant_name(
-            "ensemble_model", ensemble_key
-        )
-
-        self.assertEqual(name, (False, "ensemble_model_config_default"))
-
-    def test_ensemble_basic(self):
-        """
-        Test that we can increment the ensemble config numbers
-        """
-        sub_configA = ModelConfigVariant(ModelConfig({}), "modelA_config_0")
-        sub_configB = ModelConfigVariant(ModelConfig({}), "modelB_config_0")
-
-        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
-            [sub_configA, sub_configB]
-        )
-
-        name = self._mvnm.get_ensemble_model_variant_name(
-            "ensemble_model", ensemble_key
-        )
-
-        self.assertEqual(name, (False, "ensemble_model_config_0"))
-
-        sub_configB = ModelConfigVariant(ModelConfig({}), "modelB_config_1")
-
-        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
-            [sub_configA, sub_configB]
-        )
-
-        name = self._mvnm.get_ensemble_model_variant_name(
-            "ensemble_model", ensemble_key
-        )
-
-        self.assertEqual(name, (False, "ensemble_model_config_1"))
-
-        sub_configA = ModelConfigVariant(ModelConfig({}), "modelA_config_1")
-
-        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
-            [sub_configA, sub_configB]
-        )
-
-        name = self._mvnm.get_ensemble_model_variant_name(
-            "ensemble_model", ensemble_key
-        )
-
-        self.assertEqual(name, (False, "ensemble_model_config_2"))
-
-    def test_ensemble_repeat(self):
-        """
-        Calling with the same model name/ensemble key multiple times
-        should result in the same config name being returned
-        """
-        sub_configA = ModelConfigVariant(ModelConfig({}), "modelA_config_0")
-        sub_configB = ModelConfigVariant(ModelConfig({}), "modelB_config_0")
-
-        ensemble_key = ModelVariantNameManager.make_ensemble_composing_model_key(
-            [sub_configA, sub_configB]
-        )
-
-        name = self._mvnm.get_ensemble_model_variant_name(
-            "ensemble_model", ensemble_key
-        )
-
-        self.assertEqual(name, (False, "ensemble_model_config_0"))
-
-        name = self._mvnm.get_ensemble_model_variant_name(
-            "ensemble_model", ensemble_key
-        )
-
-        self.assertEqual(name, (True, "ensemble_model_config_0"))
-
-    def test_from_dict(self):
-        """
-        Restoring from a dict should see existing configs
-        """
-        _ = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 1}, self._non_default_param_combo
-        )
-        _ = self._mvnm.get_model_variant_name(
-            "modelB", {"A": 1}, self._non_default_param_combo
-        )
-        _ = self._mvnm.get_model_variant_name(
-            "modelA", {"A": 2}, self._non_default_param_combo
-        )
-
-        mvnm_dict = default_encode(self._mvnm)
-
-        mvnm = ModelVariantNameManager.from_dict(mvnm_dict)
-
-        self.assertEqual(mvnm._model_config_dicts, self._mvnm._model_config_dicts)
-        self.assertEqual(mvnm._model_name_index, self._mvnm._model_name_index)
-
-        a0 = mvnm.get_model_variant_name(
-            "modelA", {"A": 1}, self._non_default_param_combo
-        )
-        b0 = mvnm.get_model_variant_name(
-            "modelB", {"A": 1}, self._non_default_param_combo
-        )
-        a1 = mvnm.get_model_variant_name(
-            "modelA", {"A": 2}, self._non_default_param_combo
-        )
-        b1 = mvnm.get_model_variant_name(
-            "modelB", {"A": 2}, self._non_default_param_combo
-        )
-
-        self.assertEqual(a0, (True, "modelA_config_0"))
-        self.assertEqual(a1, (True, "modelA_config_1"))
-        self.assertEqual(b0, (True, "modelB_config_0"))
-        self.assertEqual(b1, (False, "modelB_config_1"))
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_neighborhood.py b/src/c++/perf_analyzer/tests/test_neighborhood.py
deleted file mode 100755
index ad96899fa..000000000
--- a/src/c++/perf_analyzer/tests/test_neighborhood.py
+++ /dev/null
@@ -1,916 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from typing import List
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.config.generate.coordinate import Coordinate
-from model_analyzer.config.generate.coordinate_data import CoordinateData
-from model_analyzer.config.generate.neighborhood import Neighborhood
-from model_analyzer.config.generate.search_config import NeighborhoodConfig
-from model_analyzer.config.generate.search_dimension import SearchDimension
-from model_analyzer.config.generate.search_dimensions import SearchDimensions
-
-from .common import test_result_collector as trc
-from .common.test_utils import (
-    construct_constraint_manager,
-    construct_run_config_measurement,
-)
-
-
-class TestNeighborhood(trc.TestResultCollector):
-    def setUp(self):
-        self.constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA
-            """
-        )
-
-    # Coordinates can't be sorted by default, but if we convert each one
-    # to a list those can be sorted
-    def _sort_coordinates(self, coordinates: List[Coordinate]) -> List[Coordinate]:
-        sorted_coordinates = coordinates
-        sorted_coordinates.sort(key=lambda c: [x for x in c])
-        return sorted_coordinates
-
-    def tearDown(self):
-        patch.stopall()
-
-    def _construct_rcm(self, throughput: float, latency: float):
-        model_config_name = ["modelA_config_0"]
-
-        # yapf: disable
-        non_gpu_metric_values = [{
-            "perf_throughput": throughput,
-            "perf_latency_p99": latency
-        }]
-        # yapf: enable
-
-        metric_objectives = [{"perf_throughput": 1}]
-        weights = [1]
-
-        rcm = construct_run_config_measurement(
-            model_name="modelA",
-            model_config_names=model_config_name,
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values={},
-            non_gpu_metric_values=non_gpu_metric_values,
-            constraint_manager=self.constraint_manager,
-            metric_objectives=metric_objectives,
-            model_config_weights=weights,
-        )
-        return rcm
-
-    def test_calc_distance(self):
-        a = Coordinate([1, 4, 6, 3])
-        b = Coordinate([4, 2, 6, 0])
-
-        # Euclidean distance is the square root of the
-        # sum of the distances of the coordinates
-        #
-        # Distance = sqrt( (1-4)^2 + (4-2)^2 + (6-6)^2 + (3-0)^2)
-        # Distance = sqrt( 9 + 4 + 0 + 9 )
-        # Distance = sqrt(22)
-        # Distance = 4.69
-        self.assertAlmostEqual(Neighborhood.calc_distance(a, b), 4.69, places=3)
-
-    def test_create_neighborhood(self):
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        n = Neighborhood(
-            nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=CoordinateData()
-        )
-
-        # These are all values within radius of 2 from [1,1,1]
-        # but within the bounds (no negative values)
-        #
-        expected_neighborhood = [
-            [0, 0, 0],
-            [0, 0, 1],
-            [0, 0, 2],
-            [0, 1, 0],
-            [0, 1, 1],
-            [0, 1, 2],
-            [0, 2, 0],
-            [0, 2, 1],
-            [0, 2, 2],
-            [1, 0, 0],
-            [1, 0, 1],
-            [1, 0, 2],
-            [1, 1, 0],
-            [1, 1, 1],
-            [1, 1, 2],
-            [1, 1, 3],
-            [1, 2, 0],
-            [1, 2, 1],
-            [1, 2, 2],
-            [1, 3, 1],
-            [2, 0, 0],
-            [2, 0, 1],
-            [2, 0, 2],
-            [2, 1, 0],
-            [2, 1, 1],
-            [2, 1, 2],
-            [2, 2, 0],
-            [2, 2, 1],
-            [2, 2, 2],
-            [3, 1, 1],
-        ]
-
-        expected_coordinates = [Coordinate(x) for x in expected_neighborhood]
-
-        self.assertEqual(
-            self._sort_coordinates(n._neighborhood),
-            self._sort_coordinates(expected_coordinates),
-        )
-
-    def test_large_neighborhood(self):
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("a1", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("b1", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("c1", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-        dims.add_dimensions(
-            1,
-            [
-                SearchDimension("a2", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("b2", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("c2", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-        nc = NeighborhoodConfig(dims, radius=3, min_initialized=3)
-        n = Neighborhood(
-            nc,
-            home_coordinate=Coordinate([1, 1, 1, 1, 1, 1]),
-            coordinate_data=CoordinateData(),
-        )
-
-        self.assertEqual(2328, len(n._neighborhood))
-
-    def test_num_initialized(self):
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
-
-        rcm = self._construct_rcm(throughput=100, latency=80)
-
-        # Start with 0 initialized
-        self.assertEqual(0, len(n._get_coordinates_with_valid_measurements()))
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Home coordinate is ignored. No change to num initialized/visited.
-        cd.set_measurement(Coordinate([1, 1, 1]), rcm)
-        self.assertEqual(0, len(n._get_coordinates_with_valid_measurements()))
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Incremented to 1 initialized
-        cd.set_measurement(Coordinate([0, 0, 0]), rcm)
-        self.assertEqual(1, len(n._get_coordinates_with_valid_measurements()))
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Set same point. No change to num initialized
-        cd.set_measurement(Coordinate([0, 0, 0]), rcm)
-        self.assertEqual(1, len(n._get_coordinates_with_valid_measurements()))
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Set a point outside of neighborhood
-        cd.set_measurement(Coordinate([0, 0, 4]), rcm)
-        self.assertEqual(1, len(n._get_coordinates_with_valid_measurements()))
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Set a third point inside of neighborhood
-        cd.set_measurement(Coordinate([1, 0, 0]), rcm)
-        self.assertEqual(2, len(n._get_coordinates_with_valid_measurements()))
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Set the none measurement. No changed to num initialized.
-        cd.set_measurement(Coordinate([0, 1, 1]), None)
-        self.assertEqual(2, len(n._get_coordinates_with_valid_measurements()))
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Set the last point inside of neighborhood
-        cd.set_measurement(Coordinate([1, 1, 0]), rcm)
-        self.assertEqual(3, len(n._get_coordinates_with_valid_measurements()))
-        self.assertTrue(n.enough_coordinates_initialized())
-
-    def test_get_all_adjacent_neighbors(self):
-        """
-        Test that _get_all_adjacent_neighbors() works, and understands dimension bounds
-
-        For this test, home is [0,1,4].
-
-        The possible adjacent neighbors are:
-          [-1,1,4], [1,1,4], [0,0,4], [0,2,4], [0,1,3], [0,1,5]
-
-        However, [-1,1,4] and [0,1,5] are outside of the dimension bounds and should not
-        be part of the returned list
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("a", SearchDimension.DIMENSION_TYPE_LINEAR, min=0),
-                SearchDimension("b", SearchDimension.DIMENSION_TYPE_LINEAR, min=0),
-                SearchDimension(
-                    "c", SearchDimension.DIMENSION_TYPE_LINEAR, min=1, max=4
-                ),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        n = Neighborhood(
-            nc, home_coordinate=Coordinate([0, 1, 4]), coordinate_data=CoordinateData()
-        )
-
-        adjacent_neighbors = n._get_all_adjacent_neighbors()
-        expected_list = [
-            Coordinate([1, 1, 4]),
-            Coordinate([0, 0, 4]),
-            Coordinate([0, 2, 4]),
-            Coordinate([0, 1, 3]),
-        ]
-
-        self.assertEqual(adjacent_neighbors, expected_list)
-
-    def test_num_initialized_slow_mode(self):
-        """
-        Test that _enough_coordinates_initialized() works in slow mode
-
-        Start with home=[0,0,1] in slow mode.
-        Only once all of the adjacent neighbors are added should
-        enough_coordinates_initialized() return true
-
-        Those adjacent neighbors are [1,0,1], [0,1,1], [0,0,0], [0,0,2]
-        """
-
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0, 1]), coordinate_data=cd)
-        n.force_slow_mode()
-
-        # Start with None initialized
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Home coordinate is ignored
-        cd.set_measurement(Coordinate([0, 0, 1]), MagicMock())
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Add 3 of the 4 neighbors
-        cd.set_measurement(Coordinate([1, 0, 1]), MagicMock())
-        self.assertFalse(n.enough_coordinates_initialized())
-        cd.set_measurement(Coordinate([0, 1, 1]), MagicMock())
-        self.assertFalse(n.enough_coordinates_initialized())
-        cd.set_measurement(Coordinate([0, 0, 0]), MagicMock())
-        self.assertFalse(n.enough_coordinates_initialized())
-
-        # Add the final neighbor
-        cd.set_measurement(Coordinate([0, 0, 2]), MagicMock())
-        self.assertTrue(n.enough_coordinates_initialized())
-
-    def test_is_slow_mode(self):
-        """
-        Test that _is_slow_mode() only returns true by these 3 conditions:
-        - force_slow_mode() has been called
-        - Home is passing and anyone in the neighborhood is failing
-        - Home is failing and anyone in the neighborhood is passing
-
-        Test this by running a number of sub-tests with different configurations
-        of num_passing, num_failing, home_passing, and home_measured
-        """
-
-        # Home not measured -> False
-        self._test_is_slow_mode_helper(home_measured=False, expected_result=False)
-
-        # Home passing, all passing -> False
-        self._test_is_slow_mode_helper(
-            home_passing=True, num_passing=3, num_failing=0, expected_result=False
-        )
-
-        # Same as previous, but force_slow_mode is called -> True
-        self._test_is_slow_mode_helper(
-            force_slow=True,
-            home_passing=True,
-            num_passing=3,
-            num_failing=0,
-            expected_result=True,
-        )
-
-        # Home passing, some failing -> True
-        self._test_is_slow_mode_helper(
-            num_passing=3, num_failing=1, home_passing=True, expected_result=True
-        )
-
-        # Home passing, all failing -> True
-        self._test_is_slow_mode_helper(
-            num_passing=0, num_failing=3, home_passing=True, expected_result=True
-        )
-
-        # Home failing, all failing -> False
-        self._test_is_slow_mode_helper(
-            num_passing=0, num_failing=3, home_passing=False, expected_result=False
-        )
-
-        # Home failing, some passing -> True
-        self._test_is_slow_mode_helper(
-            num_passing=1, num_failing=3, home_passing=False, expected_result=True
-        )
-
-        # Home failing, all passing -> True
-        self._test_is_slow_mode_helper(
-            num_passing=3, num_failing=0, home_passing=False, expected_result=True
-        )
-
-    def _test_is_slow_mode_helper(
-        self,
-        home_measured=True,
-        num_passing=1,
-        num_failing=1,
-        home_passing=True,
-        force_slow=False,
-        expected_result=False,
-    ):
-        total_measurements = num_failing + num_passing
-
-        passing_ret_val = [[0] * num_passing, []]
-        all_ret_val = [[0] * total_measurements, []]
-
-        patches = []
-        patches.append(
-            patch.object(
-                Neighborhood, "_is_home_measured", MagicMock(return_value=home_measured)
-            )
-        )
-        patches.append(
-            patch.object(
-                Neighborhood,
-                "_get_measurements_passing_constraints",
-                MagicMock(return_value=passing_ret_val),
-            )
-        )
-        patches.append(
-            patch.object(
-                Neighborhood,
-                "_get_all_measurements",
-                MagicMock(return_value=all_ret_val),
-            )
-        )
-        patches.append(
-            patch.object(
-                Neighborhood,
-                "_is_home_passing_constraints",
-                MagicMock(return_value=home_passing),
-            )
-        )
-
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        n = Neighborhood(
-            nc, home_coordinate=Coordinate([1, 1]), coordinate_data=CoordinateData()
-        )
-
-        for p in patches:
-            p.start()
-
-        if force_slow:
-            n.force_slow_mode()
-
-        self.assertEqual(n._is_slow_mode(), expected_result)
-
-        for p in patches:
-            p.stop()
-
-    def test_get_all_measurements(self):
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
-
-        rcm0 = self._construct_rcm(throughput=100, latency=50)
-        rcm1 = self._construct_rcm(throughput=700, latency=350)
-        rcm2 = self._construct_rcm(throughput=300, latency=130)
-
-        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
-        cd.set_measurement(Coordinate([2, 1, 1]), rcm1)
-        cd.set_measurement(Coordinate([1, 2, 1]), rcm2)
-        cd.set_measurement(Coordinate([1, 2, 2]), None)
-
-        expected_vectors = [
-            Coordinate([1, 2, 1]) - Coordinate([1, 1, 1]),
-            Coordinate([2, 1, 1]) - Coordinate([1, 1, 1]),
-        ]
-        expected_measurements = [rcm2, rcm1]
-
-        vectors, measurements = n._get_all_measurements()
-        for ev, em in zip(expected_vectors, expected_measurements):
-            self.assertIn(ev, vectors)
-            self.assertIn(em, measurements)
-
-    def test_get_constraints_passing_measurements(self):
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
-
-        # Constraints:
-        #   - Minimum throughput of 100 infer/sec
-        #   - Maximum latency of 300 ms
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 100
-                  perf_latency_p99:
-                    max: 300
-            """
-        )
-
-        rcm0 = self._construct_rcm(throughput=100, latency=50)  # pass
-        rcm0.set_constraint_manager(constraint_manager)
-
-        rcm1 = self._construct_rcm(throughput=700, latency=350)  # fail
-        rcm1.set_constraint_manager(constraint_manager)
-
-        rcm2 = self._construct_rcm(throughput=300, latency=130)  # pass
-        rcm2.set_constraint_manager(constraint_manager)
-
-        rcm3 = self._construct_rcm(throughput=400, latency=290)  # pass
-        rcm3.set_constraint_manager(constraint_manager)
-
-        rcm4 = self._construct_rcm(throughput=850, latency=400)  # fail
-        rcm4.set_constraint_manager(constraint_manager)
-
-        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
-        cd.set_measurement(Coordinate([2, 1, 1]), rcm1)
-        cd.set_measurement(Coordinate([1, 2, 1]), rcm2)
-        cd.set_measurement(Coordinate([1, 1, 2]), rcm3)
-        cd.set_measurement(Coordinate([1, 2, 2]), rcm4)
-
-        expected_vectors = [
-            Coordinate([1, 1, 2]) - Coordinate([1, 1, 1]),
-            Coordinate([1, 2, 1]) - Coordinate([1, 1, 1]),
-        ]
-        expected_measurements = [rcm3, rcm2]
-
-        vectors, measurements = n._get_measurements_passing_constraints()
-        for ev, em in zip(expected_vectors, expected_measurements):
-            self.assertIn(ev, vectors)
-            self.assertIn(em, measurements)
-
-    def test_step_vector_both_home_and_neighbors_passing_constraints(self):
-        """
-        Test _get_step_vector method where home and neighbors are passing
-
-          1. Get all vectors and measurements
-                [2, 3, 1] - [1, 1, 1] = [1, 2, 0] (throughput=300, latency=130)
-                [1, 1, 2] - [1, 1, 1] = [0, 0, 1] (throughput=400, latency=290)
-
-          2. Apply the objective-comparison weights to any non-zero dimensions
-                [1,2,0] with weight 1.0 -> [1.0, 1.0, 0.0]
-                [0,0,1] with weight 1.2 -> [0.0, 0.0, 1.2]
-
-          3. Sum the results
-                [1.0, 1.0, 0.0] + [0.0, 0.0, 1.2] = [1.0, 1.0, 1.2]
-
-          4. Divide by the sum of the vectors:
-                [1.0, 1.0, 1.2] / [1,2,1] = [1.0, 0.5, 1.2]
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=3, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
-
-        # Constraints:
-        #   - Minimum throughput of 100 infer/sec
-        #   - Maximum latency of 300 ms
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 100
-                  perf_latency_p99:
-                    max: 300
-            """
-        )
-
-        rcm0 = self._construct_rcm(throughput=100, latency=50)  # pass
-        rcm0.set_constraint_manager(constraint_manager)
-
-        rcm1 = self._construct_rcm(throughput=300, latency=130)  # pass
-        rcm1.set_constraint_manager(constraint_manager)
-
-        rcm2 = self._construct_rcm(throughput=400, latency=290)  # pass
-        rcm2.set_constraint_manager(constraint_manager)
-
-        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
-        cd.set_measurement(Coordinate([2, 3, 1]), rcm1)
-        cd.set_measurement(Coordinate([1, 1, 2]), rcm2)
-
-        hm = cd.get_measurement(Coordinate([1, 1, 1]))
-        self.assertTrue(hm.is_passing_constraints())
-
-        step_vector = n._get_step_vector()
-        expected_step_vector = Coordinate([1.0, 0.5, 1.2])
-
-        self.assertEqual(step_vector, expected_step_vector)
-
-    def test_step_vector_both_home_and_neighbors_failing_constraints(self):
-        """
-        Test _get_step_vector method where home and neighbors are failing
-
-          1. Get all vectors and measurements
-                Home latency is 450
-                [3, 2, 1] - [1, 1, 1] = [2, 1, 0] (latency=360)
-                [1, 2, 0] - [1, 1, 1] = [0, 1, -1] (latency=540)
-
-          2. Apply the constraint-comparison weights to any non-zero dimensions,
-             inverting the weight if coordinate is negative
-                [2,1,0] with weight 0.3 -> [0.3, 0.3, 0.0]
-                [0,1,-1] with weight -0.3 -> [0.0, -0.3, 0.3]
-
-          3. Sum the results
-                [0.3, 0.3, 0.0] + [0.0, -0.3, 0.3] = [0.3, 0.0, 0.3]
-
-          4. Divide by the sum of the absolute value of the vectors:
-                [0.3, 0.0, 0.3] / [2,2,1] = [0.15, 0.0, 0.3]
-
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=3, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
-
-        # Constraints:
-        #   - Minimum throughput of 100 infer/sec
-        #   - Maximum latency of 300 ms
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 100
-                  perf_latency_p99:
-                    max: 300
-            """
-        )
-
-        rcm0 = self._construct_rcm(throughput=500, latency=450)  # fail
-        rcm0.set_constraint_manager(constraint_manager)
-
-        rcm1 = self._construct_rcm(throughput=700, latency=360)  # fail
-        rcm1.set_constraint_manager(constraint_manager)
-
-        rcm2 = self._construct_rcm(throughput=850, latency=540)  # fail
-        rcm2.set_constraint_manager(constraint_manager)
-
-        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
-        cd.set_measurement(Coordinate([3, 2, 1]), rcm1)
-        cd.set_measurement(Coordinate([1, 2, 0]), rcm2)
-
-        hm = cd.get_measurement(Coordinate([1, 1, 1]))
-        self.assertFalse(hm.is_passing_constraints())
-
-        step_vector = n._get_step_vector()
-        expected_step_vector = Coordinate([0.15, 0.0, 0.3])
-        self.assertEqual(step_vector, expected_step_vector)
-
-    def test_determine_new_home_one_dimension(self):
-        """
-        Test determine_new_home for a case where only
-        one of the dimensions increases the measurement.
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0]), coordinate_data=cd)
-
-        rcm0 = self._construct_rcm(throughput=1, latency=5)
-        rcm1 = self._construct_rcm(throughput=3, latency=5)
-        rcm2 = self._construct_rcm(throughput=1, latency=5)
-
-        cd.set_measurement(Coordinate([0, 0]), rcm0)  # home coordinate
-        cd.set_measurement(Coordinate([1, 0]), rcm1)
-        cd.set_measurement(Coordinate([0, 1]), rcm2)
-
-        new_coord = n.determine_new_home()
-        self.assertEqual(new_coord, Coordinate([2, 0]))
-
-    def test_determine_new_home_two_dimensions(self):
-        """
-        Test determine_new_home for a case where both of the
-        dimensions increases the measurement equally.
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0]), coordinate_data=cd)
-
-        rcm0 = self._construct_rcm(throughput=1, latency=5)
-        rcm1 = self._construct_rcm(throughput=3, latency=5)
-        rcm2 = self._construct_rcm(throughput=3, latency=5)
-
-        cd.set_measurement(Coordinate([0, 0]), rcm0)  # home coordinate
-        cd.set_measurement(Coordinate([1, 0]), rcm1)
-        cd.set_measurement(Coordinate([0, 1]), rcm2)
-
-        new_coord = n.determine_new_home()
-        self.assertEqual(new_coord, Coordinate([2, 2]))
-
-    def test_determine_new_home(self):
-        """
-        Test determine_new_home for a case where both of the
-        dimensions increases the measurement equally
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0]), coordinate_data=cd)
-
-        rcm0 = self._construct_rcm(throughput=1, latency=5)
-        rcm1 = self._construct_rcm(throughput=7, latency=5)
-        rcm2 = self._construct_rcm(throughput=7, latency=5)
-
-        cd.set_measurement(Coordinate([0, 0]), rcm0)  # home coordinate
-        cd.set_measurement(Coordinate([1, 0]), rcm1)
-        cd.set_measurement(Coordinate([0, 1]), rcm2)
-
-        # 7x increase in throughputs will result in the maximum step of [3,3]
-        new_coord = n.determine_new_home()
-        self.assertEqual(new_coord, Coordinate([3, 3]))
-
-        new_coord = n.determine_new_home()
-        self.assertEqual(new_coord, Coordinate([3, 3]))
-
-    def test_determine_new_home_out_of_bounds(self):
-        """
-        Test that determine_new_home will clamp the result to
-        the search dimension bounds.
-
-        Both dimensions are defined to only be from 2-7. The test sets up
-        the case where the next step WOULD be from [3,6] -> [0,9] if not
-        for bounding into the defined range.
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension(
-                    "foo", SearchDimension.DIMENSION_TYPE_LINEAR, min=2, max=7
-                ),
-                SearchDimension(
-                    "bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL, min=2, max=7
-                ),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=8, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([3, 6]), coordinate_data=cd)
-
-        rcm0 = self._construct_rcm(throughput=1, latency=5)
-        rcm1 = self._construct_rcm(throughput=7, latency=5)
-
-        cd.set_measurement(Coordinate([3, 6]), rcm0)  # home coordinate
-        cd.set_measurement(Coordinate([2, 7]), rcm1)
-
-        new_coord = n.determine_new_home()
-        self.assertEqual(new_coord, Coordinate([2, 7]))
-
-    def test_all_same_throughputs(self):
-        """
-        Test that when all the coordinates in the neighborhood has the
-        same throughputs, the step vector is zero and new coordinate is
-        same as the home coordinate.
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        nc = NeighborhoodConfig(dims, radius=3, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([1, 1, 1]), coordinate_data=cd)
-
-        rcm0 = self._construct_rcm(throughput=10, latency=5)
-        rcm1 = self._construct_rcm(throughput=10, latency=5)
-        rcm2 = self._construct_rcm(throughput=10, latency=5)
-        rcm3 = self._construct_rcm(throughput=10, latency=5)
-
-        cd.set_measurement(Coordinate([1, 1, 1]), rcm0)  # home coordinate
-        cd.set_measurement(Coordinate([1, 0, 0]), rcm1)
-        cd.set_measurement(Coordinate([0, 1, 0]), rcm2)
-        cd.set_measurement(Coordinate([0, 0, 1]), rcm3)
-
-        step_vector = n._get_step_vector()
-        self.assertEqual(step_vector, Coordinate([0, 0, 0]))
-
-        n.force_slow_mode()
-        new_coord = n.determine_new_home()
-        self.assertEqual(new_coord, Coordinate([1, 1, 1]))
-
-    def test_translate_step_vector(self):
-        """
-        Test the functionality of translate_step_vector()
-
-        Any dimension value that is less than translation_list[0] will get 0
-
-        Any dimension value that is between translation_list[i] and translation_list[i+1]
-        will get i+1
-        """
-
-        # Test generic
-        self._test_translate_step_vector_helper(
-            input_vector=[0.35, 0.25, 0.15, 0.05],
-            translation_list=[0.1, 0.2, 0.3],
-            expected_vector=[3, 2, 1, 0],
-        )
-
-        # Test negative
-        self._test_translate_step_vector_helper(
-            input_vector=[-0.35, -0.25, -0.15, -0.05],
-            translation_list=[0.1, 0.2, 0.3],
-            expected_vector=[-3, -2, -1, 0],
-        )
-
-        # Test boundary case. Exact match rounds down
-        self._test_translate_step_vector_helper(
-            input_vector=[0.099, 0.100, 0.101],
-            translation_list=[0.1],
-            expected_vector=[0, 0, 1],
-        )
-
-    def _test_translate_step_vector_helper(
-        self, input_vector, translation_list, expected_vector
-    ):
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0, [SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR)]
-        )
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([0]), coordinate_data=cd)
-
-        input_coord = Coordinate(input_vector)
-        expected_coord = Coordinate(expected_vector)
-        output_coord = n._translate_step_vector(input_coord, translation_list)
-
-        self.assertEqual(expected_coord, output_coord)
-
-    def _test_calculate_step_vector_from_vectors_and_weights(self):
-        """
-        Test the functionality of __test_calculate_step_vector_from_vectors_and_weights()
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-                SearchDimension("foobar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-        nc = NeighborhoodConfig(dims, radius=2, min_initialized=3)
-        cd = CoordinateData()
-        n = Neighborhood(nc, home_coordinate=Coordinate([0, 0, 0]), coordinate_data=cd)
-
-        vectors = [[0, 0, 1], [1, -3, 2], [-4, 2, 0], [3, 2, 6]]
-        weights = [0.5, -3.0, -1, 0]
-
-        # Sum up weights into vectors:
-        # [0,0,0.5] + [-3,3,-3] + [1, -1, 0] + [0,0,0] = [-2, 2, -2.5]
-        #
-        # Sum up absolute value of vectors:
-        # [0,0,1] + [1,3,2] + [4,2,0] + [3,2,6] = [8, 7, 9]
-        #
-        # Divide:
-        # [-2, 2, -2.5] / [8, 7, 9] = [-2/8, 2/7, -2.5/9]
-
-        step_vector = n._calculate_step_vector_from_vectors_and_weights(
-            vectors, weights
-        )
-        expected_result = [-2 / 8, 2 / 7, -2.5 / 9]
-        self.assertEqual(step_vector, expected_result)
diff --git a/src/c++/perf_analyzer/tests/test_parameter_search.py b/src/c++/perf_analyzer/tests/test_parameter_search.py
deleted file mode 100755
index 7f410bb26..000000000
--- a/src/c++/perf_analyzer/tests/test_parameter_search.py
+++ /dev/null
@@ -1,336 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from math import log2
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.config.input.config_command_profile import ConfigCommandProfile
-from model_analyzer.config.input.config_defaults import (
-    DEFAULT_RUN_CONFIG_MAX_CONCURRENCY,
-    DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE,
-    DEFAULT_RUN_CONFIG_MIN_CONCURRENCY,
-    DEFAULT_RUN_CONFIG_MIN_REQUEST_RATE,
-)
-from model_analyzer.constants import THROUGHPUT_MINIMUM_CONSECUTIVE_PARAMETER_TRIES
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.result.constraint_manager import ConstraintManager
-from model_analyzer.result.parameter_search import ParameterSearch
-from model_analyzer.result.run_config_measurement import RunConfigMeasurement
-
-from .common import test_result_collector as trc
-from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
-
-
-class TestParameterSearch(trc.TestResultCollector):
-    def setUp(self):
-        self._min_concurrency_index = int(log2(DEFAULT_RUN_CONFIG_MIN_CONCURRENCY))
-        self._max_concurrency_index = int(log2(DEFAULT_RUN_CONFIG_MAX_CONCURRENCY))
-
-        self._expected_concurrencies = [
-            2**c
-            for c in range(self._min_concurrency_index, self._max_concurrency_index + 1)
-        ]
-        self._concurrencies = []
-
-        self._min_request_rate_index = int(log2(DEFAULT_RUN_CONFIG_MIN_REQUEST_RATE))
-        self._max_request_rate_index = int(log2(DEFAULT_RUN_CONFIG_MAX_REQUEST_RATE))
-
-        self._expected_request_rates = [
-            2**rr
-            for rr in range(
-                self._min_request_rate_index, self._max_request_rate_index + 1
-            )
-        ]
-        self._request_rates = []
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_concurrency_sweep(self):
-        """
-        Test sweeping concurrency from min to max, when no constraints are present
-        and throughput is linearly increasing
-        """
-        config = self._create_single_model_no_constraints()
-        constraint_manager = ConstraintManager(config)
-        concurrency_search = ParameterSearch(config)
-
-        for concurrency in concurrency_search.search_parameters():
-            self._concurrencies.append(concurrency)
-
-            concurrency_search.add_run_config_measurement(
-                run_config_measurement=self._construct_rcm(
-                    throughput=100 * concurrency,
-                    latency=10,
-                    concurrency=concurrency,
-                    constraint_manager=constraint_manager,
-                )
-            )
-
-        self.assertEqual(self._concurrencies, self._expected_concurrencies)
-
-    def test_request_rate_sweep(self):
-        """
-        Test sweeping request rate from min to max, when no constraints are present
-        and throughput is linearly increasing
-        """
-        config = self._create_single_model_no_constraints()
-        constraint_manager = ConstraintManager(config)
-        concurrency_search = ParameterSearch(
-            config, model_parameters={"request_rate": "True"}
-        )
-
-        for request_rate in concurrency_search.search_parameters():
-            self._request_rates.append(request_rate)
-
-            concurrency_search.add_run_config_measurement(
-                run_config_measurement=self._construct_rcm(
-                    throughput=100 * request_rate,
-                    latency=10,
-                    request_rate=request_rate,
-                    constraint_manager=constraint_manager,
-                )
-            )
-
-        self.assertEqual(self._request_rates, self._expected_request_rates)
-
-    def test_saturating_sweep(self):
-        """
-        Test sweeping concurrency from min to max, when no constraints are present
-        and throughput increases and then saturates
-        """
-        config = self._create_single_model_no_constraints()
-        constraint_manager = ConstraintManager(config)
-        concurrency_search = ParameterSearch(config)
-        INCREASE_THROUGHPUT_COUNT = 4
-
-        # [100, 200, 400, 800, 1000, 1000,...]
-        throughputs = [
-            100 * 2**c if c < INCREASE_THROUGHPUT_COUNT else 1000
-            for c in range(self._min_concurrency_index, self._max_concurrency_index + 1)
-        ]
-
-        for i, concurrency in enumerate(concurrency_search.search_parameters()):
-            self._concurrencies.append(concurrency)
-
-            concurrency_search.add_run_config_measurement(
-                run_config_measurement=self._construct_rcm(
-                    throughput=throughputs[i],
-                    latency=10,
-                    concurrency=concurrency,
-                    constraint_manager=constraint_manager,
-                )
-            )
-
-        expected_concurrencies = [
-            2**c
-            for c in range(
-                INCREASE_THROUGHPUT_COUNT
-                + THROUGHPUT_MINIMUM_CONSECUTIVE_PARAMETER_TRIES
-            )
-        ]
-        self.assertEqual(self._concurrencies, expected_concurrencies)
-
-    def test_sweep_with_constraints_decreasing(self):
-        """
-        Test sweeping concurrency from min to max, with 95ms latency constraint
-        and throughput is linearly increasing - which causes a decreasing binary search
-        """
-        config = self._create_single_model_with_constraints("95")
-        constraint_manager = ConstraintManager(config)
-        concurrency_search = ParameterSearch(config)
-
-        self._expected_concurrencies.extend([12, 10, 9])
-        latencies = [10 * c for c in self._expected_concurrencies]
-
-        for i, concurrency in enumerate(concurrency_search.search_parameters()):
-            self._concurrencies.append(concurrency)
-
-            concurrency_search.add_run_config_measurement(
-                run_config_measurement=self._construct_rcm(
-                    throughput=100 * concurrency,
-                    latency=latencies[i],
-                    concurrency=concurrency,
-                    constraint_manager=constraint_manager,
-                )
-            )
-
-        self.assertEqual(self._concurrencies, self._expected_concurrencies)
-
-    def test_sweep_with_constraints_decrease_then_increase(self):
-        """
-        Test sweeping concurrency from min to max, with 155ms latency constraint
-        and throughput is linearly increasing - which causes a decreasing, then increasing binary search
-        """
-        config = self._create_single_model_with_constraints("155")
-        constraint_manager = ConstraintManager(config)
-        concurrency_search = ParameterSearch(config)
-
-        self._expected_concurrencies.extend([12, 14, 15])
-        latencies = [10 * c for c in self._expected_concurrencies]
-
-        for i, concurrency in enumerate(concurrency_search.search_parameters()):
-            self._concurrencies.append(concurrency)
-
-            concurrency_search.add_run_config_measurement(
-                run_config_measurement=self._construct_rcm(
-                    throughput=100 * concurrency,
-                    latency=latencies[i],
-                    concurrency=concurrency,
-                    constraint_manager=constraint_manager,
-                )
-            )
-
-        self.assertEqual(self._concurrencies, self._expected_concurrencies)
-
-    def test_sweep_with_multiple_violation_areas(self):
-        """
-        Test sweeping concurrency from min to max, with 155ms latency constraint
-        with violations in two separate locations
-        """
-        config = self._create_single_model_with_constraints("155")
-        constraint_manager = ConstraintManager(config)
-        concurrency_search = ParameterSearch(config)
-
-        self._expected_concurrencies.extend([12, 14, 15])
-        latencies = [10 * c for c in self._expected_concurrencies]
-        # this adds an early constraint violation which should be ignored
-        latencies[1] = 200
-
-        for i, concurrency in enumerate(concurrency_search.search_parameters()):
-            self._concurrencies.append(concurrency)
-
-            concurrency_search.add_run_config_measurement(
-                run_config_measurement=self._construct_rcm(
-                    throughput=100 * concurrency,
-                    latency=latencies[i],
-                    concurrency=concurrency,
-                    constraint_manager=constraint_manager,
-                )
-            )
-
-        self.assertEqual(self._concurrencies, self._expected_concurrencies)
-
-    def test_sweep_with_constraints_hitting_limit(self):
-        """
-        Test sweeping concurrency from min to max, with 970ms latency constraint
-        and throughput matches concurrency, this will cause BCS to
-        quit after DEFAULT_RUN_CONFIG_MAX_BINARY_SEARCH_STEPS (5) attempts
-        """
-        config = self._create_single_model_with_constraints("970")
-        constraint_manager = ConstraintManager(config)
-        concurrency_search = ParameterSearch(config)
-
-        self._expected_concurrencies.extend([768, 896, 960, 992, 976])
-        latencies = self._expected_concurrencies
-
-        for i, concurrency in enumerate(concurrency_search.search_parameters()):
-            self._concurrencies.append(concurrency)
-
-            concurrency_search.add_run_config_measurement(
-                run_config_measurement=self._construct_rcm(
-                    throughput=100 * concurrency,
-                    latency=latencies[i],
-                    concurrency=concurrency,
-                    constraint_manager=constraint_manager,
-                )
-            )
-
-        self.assertEqual(self._concurrencies, self._expected_concurrencies)
-
-    def test_not_adding_measurements(self):
-        """
-        Test that an exception is raised if measurements are not added
-        """
-        config = self._create_single_model_no_constraints()
-        constraint_manager = ConstraintManager(config)
-        concurrency_search = ParameterSearch(config)
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            for concurrency in concurrency_search.search_parameters():
-                self._concurrencies.append(concurrency)
-
-                if concurrency < 32:
-                    concurrency_search.add_run_config_measurement(
-                        run_config_measurement=self._construct_rcm(
-                            throughput=100 * concurrency,
-                            latency=10,
-                            concurrency=concurrency,
-                            constraint_manager=constraint_manager,
-                        )
-                    )
-
-    def _create_single_model_no_constraints(self):
-        args = ["model-analyzer", "profile", "--profile-models", "test_model"]
-        yaml_str = ""
-        config = evaluate_mock_config(args, yaml_str)
-
-        return config
-
-    def _create_single_model_with_constraints(
-        self, latency_budget: str
-    ) -> ConfigCommandProfile:
-        args = [
-            "model-analyzer",
-            "profile",
-            "--profile-models",
-            "test_model",
-            "--latency-budget",
-            latency_budget,
-        ]
-        yaml_str = ""
-        config = evaluate_mock_config(args, yaml_str)
-
-        return config
-
-    def _construct_rcm(
-        self,
-        throughput: int,
-        latency: int,
-        constraint_manager: ConstraintManager,
-        concurrency: int = 0,
-        request_rate: int = 0,
-    ) -> RunConfigMeasurement:
-        if concurrency:
-            self.model_specific_pa_params = [
-                {"batch_size": 1, "concurrency": concurrency}
-            ]
-        else:
-            self.model_specific_pa_params = [
-                {"batch_size": 1, "request_rate": request_rate}
-            ]
-
-        self.rcm0_non_gpu_metric_values = [
-            {
-                "perf_throughput": throughput,
-                "perf_latency_p99": latency,
-                "cpu_used_ram": 1000,
-            }
-        ]
-
-        return construct_run_config_measurement(
-            model_name="test_model",
-            model_config_names=["test_model_config_0"],
-            model_specific_pa_params=self.model_specific_pa_params,
-            gpu_metric_values=MagicMock(),
-            non_gpu_metric_values=self.rcm0_non_gpu_metric_values,
-            constraint_manager=constraint_manager,
-        )
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_pdf_report.py b/src/c++/perf_analyzer/tests/test_pdf_report.py
deleted file mode 100755
index acd938aed..000000000
--- a/src/c++/perf_analyzer/tests/test_pdf_report.py
+++ /dev/null
@@ -1,61 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.reports.pdf_report import PDFReport
-
-from .common import test_result_collector as trc
-
-
-class TestPDFReportMethods(trc.TestResultCollector):
-    """
-    Tests the methods of the PDFReport class
-    """
-
-    def setUp(self):
-        self.maxDiff = None
-        self.report = PDFReport()
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_write_report(self):
-        with patch(
-            "model_analyzer.reports.pdf_report.pdfkit", MagicMock()
-        ) as pdfkit_mock:
-            self.report.add_title("Test PDF Report")
-            self.report.add_subheading("Throughput vs. Latency")
-            test_paragraph = (
-                "This is a test paragraph with a lot to say."
-                " There is more than one line in this paragraph."
-            )
-            self.report.add_paragraph(test_paragraph, font_size=14)
-            self.report.write_report("test_report_filename")
-
-            expected_report_body = (
-                "<html><head><style></style></head><body><center><h1>Test PDF Report</h1></center><h3>Throughput vs. Latency</h3>"
-                f'<div style="font-size:14"><p>{test_paragraph}</p></div>'
-                "</body></html>"
-            )
-            pdfkit_mock.from_string.assert_called_with(
-                expected_report_body, "test_report_filename", options={"quiet": ""}
-            )
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_plot.py b/src/c++/perf_analyzer/tests/test_plot.py
deleted file mode 100755
index c9e3d84c0..000000000
--- a/src/c++/perf_analyzer/tests/test_plot.py
+++ /dev/null
@@ -1,131 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.plots.simple_plot import SimplePlot
-
-from .common import test_result_collector as trc
-from .common.test_utils import construct_run_config_measurement
-from .mocks.mock_matplotlib import MockMatplotlibMethods
-
-
-class TestPlotMethods(trc.TestResultCollector):
-    def setUp(self):
-        # mocks
-        self.matplotlib_mock = MockMatplotlibMethods()
-        self.matplotlib_mock.start()
-
-    def test_create_plot(self):
-        # Create a plot and check for call to subplots
-        SimplePlot(
-            name="test_plot",
-            title="test_title",
-            x_axis="perf_throughput",
-            y_axis="perf_latency_p99",
-        )
-
-        self.matplotlib_mock.assert_called_subplots()
-
-    def test_add_measurement(self):
-        plot = SimplePlot(
-            name="test_plot",
-            title="test_title",
-            x_axis="perf_throughput",
-            y_axis="perf_latency_p99",
-        )
-
-        gpu_metric_values = {0: {"gpu_used_memory": 5000, "gpu_utilization": 50}}
-
-        non_gpu_metric_values = {"perf_throughput": 200, "perf_latency_p99": 8000}
-
-        objective_spec = {"perf_throughput": 10, "perf_latency_p99": 5}
-
-        measurement = construct_run_config_measurement(
-            model_name="test_model",
-            model_config_names=["test_model_config_0"],
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values=gpu_metric_values,
-            non_gpu_metric_values=[non_gpu_metric_values],
-            metric_objectives=[objective_spec],
-        )
-
-        # Add above measurement
-        plot.add_run_config_measurement("test_model_label1", measurement)
-        self.assertDictEqual(
-            plot.data(), {"test_model_label1": {"x_data": [200], "y_data": [8000]}}
-        )
-
-        # Add measurement again with different label
-        plot.add_run_config_measurement("test_model_label2", measurement)
-        self.assertDictEqual(
-            plot.data(),
-            {
-                "test_model_label1": {"x_data": [200], "y_data": [8000]},
-                "test_model_label2": {"x_data": [200], "y_data": [8000]},
-            },
-        )
-
-    def test_plot_data(self):
-        plot = SimplePlot(
-            name="test_plot",
-            title="test_title",
-            x_axis="perf_throughput",
-            y_axis="perf_latency_p99",
-        )
-
-        gpu_metric_values = {0: {"gpu_used_memory": 5000, "gpu_utilization": 50}}
-
-        non_gpu_metric_values = {"perf_throughput": 200, "perf_latency_p99": 8000}
-
-        objective_spec = {"perf_throughput": 10, "perf_latency_p99": 5}
-
-        measurement = construct_run_config_measurement(
-            model_name="test_model",
-            model_config_names=["test_model_config_0"],
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values=gpu_metric_values,
-            non_gpu_metric_values=[non_gpu_metric_values],
-            metric_objectives=[objective_spec],
-        )
-
-        plot.add_run_config_measurement("test_model_label", measurement)
-
-        # Call plot and assert args
-        plot.plot_data_and_constraints(constraints={})
-        self.matplotlib_mock.assert_called_plot_with_args(
-            x_data=[200], y_data=[8000], marker="o", label="test_model_label"
-        )
-
-    def test_save(self):
-        plot = SimplePlot(
-            name="test_plot",
-            title="test_title",
-            x_axis="perf_throughput",
-            y_axis="perf_latency_p99",
-        )
-
-        plot.save("test_path")
-        self.matplotlib_mock.assert_called_save_with_args("test_path/test_plot")
-
-    def tearDown(self):
-        self.matplotlib_mock.stop()
-        patch.stopall()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_plot_manager.py b/src/c++/perf_analyzer/tests/test_plot_manager.py
deleted file mode 100755
index a8e5aa3e4..000000000
--- a/src/c++/perf_analyzer/tests/test_plot_manager.py
+++ /dev/null
@@ -1,196 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import json
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.plots.plot_manager import PlotManager
-from model_analyzer.result.constraint_manager import ConstraintManager
-from model_analyzer.result.result_manager import ResultManager
-from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
-
-from .common import test_result_collector as trc
-from .common.test_utils import (
-    ROOT_DIR,
-    convert_to_bytes,
-    default_encode,
-    evaluate_mock_config,
-)
-
-
-class TestPlotManager(trc.TestResultCollector):
-    def setUp(self):
-        self.create_new_dump = False
-        self._create_single_model_result_manager()
-        self._create_multi_model_result_manager()
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_single_model_summary_plots_against_golden(self):
-        """
-        Match the summary plots against the golden versions in
-        tests/common/single-model-ckpt
-        """
-        plot_manager = PlotManager(
-            config=self._single_model_config,
-            result_manager=self._single_model_result_manager,
-            constraint_manager=MagicMock(),
-        )
-
-        plot_manager.create_summary_plots()
-
-        # Set self.create_new_dump to True for creating a new output dump to compare against.
-        # Default value is False.
-        if self.create_new_dump:
-            with open(f"{ROOT_DIR}/single-model-ckpt/plot_manager.json", "wb") as f:
-                f.write(
-                    convert_to_bytes(
-                        json.dumps(
-                            self._plot_manager_to_dict(plot_manager),
-                            default=default_encode,
-                        )
-                    )
-                )
-
-        with open(f"{ROOT_DIR}/single-model-ckpt/plot_manager.json", "rb") as f:
-            golden_plot_manager_dict = json.load(f)
-
-        plot_manager_dict = self._plot_manager_to_dict(plot_manager)
-
-        self.assertEqual(golden_plot_manager_dict, plot_manager_dict)
-
-    def test_multi_model_summary_plots_against_golden(self):
-        """
-        Match the summary plots against the golden versions in
-        tests/common/multi-model-ckpt
-        """
-        plot_manager = PlotManager(
-            config=self._multi_model_config,
-            result_manager=self._multi_model_result_manager,
-            constraint_manager=MagicMock(),
-        )
-
-        plot_manager.create_summary_plots()
-
-        # Set self.create_new_dump to True for creating a new output dump to compare against.
-        # Default value is False.
-        if self.create_new_dump:
-            with open(f"{ROOT_DIR}/multi-model-ckpt/plot_manager.json", "wb") as f:
-                f.write(
-                    convert_to_bytes(
-                        json.dumps(
-                            self._plot_manager_to_dict(plot_manager),
-                            default=default_encode,
-                        )
-                    )
-                )
-
-        with open(f"{ROOT_DIR}/multi-model-ckpt/plot_manager.json", "rb") as f:
-            golden_plot_manager_dict = json.load(f)
-
-        plot_manager_dict = self._plot_manager_to_dict(plot_manager)
-
-        self.assertEqual(golden_plot_manager_dict, plot_manager_dict)
-
-    def _create_single_model_result_manager(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "-f",
-            "config.yml",
-            "--checkpoint-directory",
-            f"{ROOT_DIR}/single-model-ckpt/",
-            "--export-path",
-            f"{ROOT_DIR}/single-model-ckpt/",
-            "--model-repository",
-            ".",
-        ]
-        yaml_str = """
-            profile_models: add_sub
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-        state_manager = AnalyzerStateManager(config=config, server=None)
-        state_manager.load_checkpoint(checkpoint_required=True)
-
-        constraint_manager = ConstraintManager(config)
-        self._single_model_result_manager = ResultManager(
-            config=config,
-            state_manager=state_manager,
-            constraint_manager=constraint_manager,
-        )
-
-        self._single_model_config = config
-
-    def _create_multi_model_result_manager(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "-f",
-            "config.yml",
-            "--checkpoint-directory",
-            f"{ROOT_DIR}/multi-model-ckpt/",
-            "--export-path",
-            f"{ROOT_DIR}/multi-model-ckpt/",
-            "--model-repository",
-            ".",
-            "--run-config-profile-models-concurrently-enable",
-            "--run-config-search-mode",
-            "quick",
-        ]
-        yaml_str = """
-            profile_models: resnet50_libtorch,vgg19_libtorch
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-        state_manager = AnalyzerStateManager(config=config, server=None)
-        state_manager.load_checkpoint(checkpoint_required=True)
-
-        constraint_manager = ConstraintManager(config)
-        self._multi_model_result_manager = ResultManager(
-            config=config,
-            state_manager=state_manager,
-            constraint_manager=constraint_manager,
-        )
-
-        self._multi_model_config = config
-
-    def _plot_manager_to_dict(self, plot_manager):
-        plot_manager_dict = {}
-        plot_manager_dict["_simple_plots"] = {}
-        for spd_key, simple_plot_dict in plot_manager._simple_plots.items():
-            plot_manager_dict["_simple_plots"][spd_key] = {}
-            for sp_key, simple_plot in simple_plot_dict.items():
-                plot_manager_dict["_simple_plots"][spd_key][
-                    sp_key
-                ] = self._simple_plot_to_dict(simple_plot)
-
-        return plot_manager_dict
-
-    def _simple_plot_to_dict(self, simple_plot):
-        simple_plot_dict = {}
-        simple_plot_dict["_name"] = simple_plot._name
-        simple_plot_dict["_title"] = simple_plot._title
-        simple_plot_dict["_x_axis"] = simple_plot._x_axis
-        simple_plot_dict["_y_axis"] = simple_plot._y_axis
-        simple_plot_dict["_monotonic"] = simple_plot._monotonic
-        simple_plot_dict["_data"] = simple_plot._data
-
-        return simple_plot_dict
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_quick_run_config_generator.py b/src/c++/perf_analyzer/tests/test_quick_run_config_generator.py
deleted file mode 100755
index 7eef2514f..000000000
--- a/src/c++/perf_analyzer/tests/test_quick_run_config_generator.py
+++ /dev/null
@@ -1,1359 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.config.generate.coordinate import Coordinate
-from model_analyzer.config.generate.model_profile_spec import ModelProfileSpec
-from model_analyzer.config.generate.model_variant_name_manager import (
-    ModelVariantNameManager,
-)
-from model_analyzer.config.generate.quick_run_config_generator import (
-    QuickRunConfigGenerator,
-)
-from model_analyzer.config.generate.run_config_generator_factory import (
-    RunConfigGeneratorFactory,
-)
-from model_analyzer.config.generate.search_config import SearchConfig
-from model_analyzer.config.generate.search_dimension import SearchDimension
-from model_analyzer.config.generate.search_dimensions import SearchDimensions
-from model_analyzer.config.input.objects.config_model_profile_spec import (
-    ConfigModelProfileSpec,
-)
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.triton.model.model_config import ModelConfig
-from tests.common.test_utils import evaluate_mock_config
-
-from .common import test_result_collector as trc
-
-
-def mock_ensemble_configs(*args, **kwargs):
-    fake_config = {
-        "name": "my-model",
-        "platform": "ensemble",
-        "ensemble_scheduling": {
-            "step": [{"model_name": "fake_model_A"}, {"model_name": "fake_model_B"}]
-        },
-        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-        "max_batch_size": 4,
-    }
-    fake_base_composing_config0 = {
-        "name": "fake_model_A",
-        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-        "max_batch_size": 4,
-        "sequence_batching": {},
-    }
-    fake_base_composing_config1 = {
-        "name": "fake_model_B",
-        "input": [{"name": "INPUT__2", "dataType": "TYPE_FP16", "dims": [32]}],
-        "max_batch_size": 8,
-    }
-
-    if args:
-        model_name = args[4]
-    else:
-        model_name = kwargs["model_name"]
-
-    if model_name == "my-model":
-        return fake_config
-    elif model_name == "fake_model_A":
-        return fake_base_composing_config0
-    elif model_name == "fake_model_B":
-        return fake_base_composing_config1
-
-    return None
-
-
-def mock_composing_ensemble_configs(*args, **kwargs):
-    fake_config = {
-        "name": "my-model",
-        "platform": "ensemble",
-        "ensemble_scheduling": {
-            "step": [{"model_name": "fake_model_A"}, {"model_name": "fake_model_B"}]
-        },
-        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-        "max_batch_size": 4,
-    }
-    fake_base_composing_config0 = {
-        "name": "fake_model_A",
-        "platform": "ensemble",
-        "ensemble_scheduling": {
-            "step": [{"model_name": "fake_model_C"}, {"model_name": "fake_model_D"}]
-        },
-        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-        "max_batch_size": 4,
-        "sequence_batching": {},
-    }
-
-    if args:
-        model_name = args[4]
-    else:
-        model_name = kwargs["model_name"]
-
-    if model_name == "my-model":
-        return fake_config
-    else:
-        return fake_base_composing_config0
-
-
-def mock_bls_configs(*args, **kwargs):
-    fake_config = {
-        "name": "my-model",
-        "platform": "bls",
-        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-        "max_batch_size": 4,
-    }
-    fake_base_composing_config0 = {
-        "name": "fake_model_A",
-        "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-        "max_batch_size": 4,
-        "sequence_batching": {},
-    }
-    fake_base_composing_config1 = {
-        "name": "fake_model_B",
-        "input": [{"name": "INPUT__2", "dataType": "TYPE_FP16", "dims": [32]}],
-        "max_batch_size": 8,
-    }
-
-    if args:
-        model_name = args[4]
-    else:
-        model_name = kwargs["model_name"]
-
-    if model_name == "my-model":
-        return fake_config
-    elif model_name == "fake_model_A":
-        return fake_base_composing_config0
-    else:
-        return fake_base_composing_config1
-
-
-class TestQuickRunConfigGenerator(trc.TestResultCollector):
-    def setUp(self):
-        fake_config = {
-            "name": "fake_model_name",
-            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-            "max_batch_size": 4,
-        }
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            return_value=fake_config,
-        ):
-            self._mock_models = [
-                ModelProfileSpec(
-                    ConfigModelProfileSpec(model_name="fake_model_name"),
-                    MagicMock(),
-                    MagicMock(),
-                    MagicMock(),
-                )
-            ]
-
-        self._dims = SearchDimensions()
-        self._dims.add_dimensions(
-            0,
-            [
-                SearchDimension(
-                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-                SearchDimension(
-                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
-                ),
-                SearchDimension(
-                    "concurrency", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-            ],
-        )
-
-        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
-        config = self._create_config()
-        self._qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            self._mock_models,
-            {},
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-    def test_get_starting_coordinate(self):
-        """Test that get_starting_coordinate() works for non-zero values"""
-        # yapf: disable
-        dims = SearchDimensions()
-        dims.add_dimensions(0, [
-                SearchDimension("x", SearchDimension.DIMENSION_TYPE_EXPONENTIAL, min=2),
-                SearchDimension("y", SearchDimension.DIMENSION_TYPE_LINEAR, min=1),
-                SearchDimension("z", SearchDimension.DIMENSION_TYPE_EXPONENTIAL, min=3)
-        ])
-        sc = SearchConfig(dimensions=dims,radius=2, min_initialized=2)
-        # yapf: enable
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            MagicMock(),
-            MagicMock(),
-            self._mock_models,
-            {},
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-        self.assertEqual(qrcg._get_starting_coordinate(), Coordinate([2, 1, 3]))
-
-    def test_get_next_run_config(self):
-        """
-        Test that get_next_run_config() creates a proper RunConfig
-
-        Sets up a case where the coordinate is [5,7], which corresponds to
-          - max_batch_size = 32
-          - instance_count = 8
-          - concurrency = 32*8*2 = 512
-
-        Also
-        - dynamic batching should be on
-        - existing values from the base model config should persist if they aren't overwritten
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "/tmp",
-            "--config-file",
-            "/tmp/my_config.yml",
-        ]
-
-        yaml_str = """
-        run_config_search_mode: quick
-        profile_models:
-          fake_model_name:
-            model_config_parameters:
-              dynamic_batching: { "preferred_batch_size": [4] }
-              optimization:
-                execution_accelerators:
-                  gpu_execution_accelerator:
-                    - name: tensorrt
-                      parameters:
-                        precision_mode: FP16
-        """
-
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        fake_config = {
-            "name": "fake_model_name",
-            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-            "max_batch_size": 4,
-        }
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            return_value=fake_config,
-        ):
-            mock_models = [
-                ModelProfileSpec(
-                    config.get_config()["profile_models"].value()[0],
-                    config,
-                    MagicMock(),
-                    MagicMock(),
-                )
-            ]
-
-        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
-
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            mock_models,
-            {},
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        qrcg._coordinate_to_measure = Coordinate([5, 7])
-
-        expected_model_config = {
-            "dynamicBatching": {},
-            "instanceGroup": [
-                {
-                    "count": 8,
-                    "kind": "KIND_GPU",
-                }
-            ],
-            "maxBatchSize": 32,
-            "name": "fake_model_name",
-            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
-            "optimization": {
-                "executionAccelerators": {
-                    "gpuExecutionAccelerator": [
-                        {"name": "tensorrt", "parameters": {"precision_mode": "FP16"}}
-                    ]
-                }
-            },
-        }
-        # yapf: enable
-
-        rc = qrcg._get_next_run_config()
-
-        self.assertEqual(len(rc.model_run_configs()), 1)
-        model_config = rc.model_run_configs()[0].model_config()
-        perf_config = rc.model_run_configs()[0].perf_config()
-
-        self.assertEqual(model_config.to_dict(), expected_model_config)
-        self.assertEqual(perf_config["concurrency-range"], 512)
-        self.assertEqual(perf_config["batch-size"], 1)
-
-    def test_get_next_run_config_multi_model(self):
-        """
-        Test that get_next_run_config() creates a proper RunConfig for multi-model
-
-        Sets up a case where the coordinate is [1,2,4,5], which corresponds to
-          - model 1 max_batch_size = 2
-          - model 1 instance_count = 3
-          - model 1 concurrency = 2*3*2 = 12
-          - model 2 max_batch_size = 16
-          - model 2 instance_count = 6
-          - model 2 concurrency = 16*6*2 = 192
-
-        Also,
-        - sequence batching should be on for model 1
-        - dynamic batching should be on for model 2
-        - existing values from the base model config should persist if they aren't overwritten
-        - existing values for perf-analyzer config should persist if they aren't overwritten
-        """
-
-        # yapf: disable
-        fake_base_config1 = {
-            "name": "fake_model_name1",
-            "input": [{
-                "name": "INPUT__0",
-                "dataType": "TYPE_FP32",
-                "dims": [16]
-            }],
-            "max_batch_size": 4,
-            "sequence_batching": {}
-        }
-        fake_base_config2 = {
-            "name": "fake_model_name2",
-            "input": [{
-                "name": "INPUT__2",
-                "dataType": "TYPE_FP16",
-                "dims": [32]
-            }],
-            "max_batch_size": 8
-        }
-
-        expected_model_config1 = {
-            'instanceGroup': [{
-                'count': 3,
-                'kind': 'KIND_GPU',
-            }],
-            'maxBatchSize': 2,
-            'sequenceBatching': {},
-            'name': 'fake_model_name1',
-            'input': [{
-                "name": "INPUT__0",
-                "dataType": "TYPE_FP32",
-                "dims": ['16']
-            }]
-        }
-
-        expected_model_config2 = {
-            'dynamicBatching': {},
-            'instanceGroup': [{
-                'count': 6,
-                'kind': 'KIND_GPU',
-            }],
-            'maxBatchSize': 16,
-            'name': 'fake_model_name2',
-            'input': [{
-                "name": "INPUT__2",
-                "dataType": "TYPE_FP16",
-                "dims": ['32']
-            }]
-        }
-        # yapf: enable
-
-        mock_models = []
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            return_value=fake_base_config1,
-        ):
-            mock_models.append(
-                ModelProfileSpec(
-                    ConfigModelProfileSpec(
-                        model_name="fake_model_name1",
-                        perf_analyzer_flags={"model-version": 2},
-                    ),
-                    MagicMock(),
-                    MagicMock(),
-                    MagicMock(),
-                )
-            )
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            return_value=fake_base_config2,
-        ):
-            mock_models.append(
-                ModelProfileSpec(
-                    ConfigModelProfileSpec(
-                        model_name="fake_model_name2",
-                        perf_analyzer_flags={"model-version": 3},
-                    ),
-                    MagicMock(),
-                    MagicMock(),
-                    MagicMock(),
-                )
-            )
-
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension(
-                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-                SearchDimension(
-                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
-                ),
-            ],
-        )
-        dims.add_dimensions(
-            1,
-            [
-                SearchDimension(
-                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-                SearchDimension(
-                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
-                ),
-            ],
-        )
-
-        sc = SearchConfig(dimensions=dims, radius=5, min_initialized=2)
-        config = self._create_config()
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            mock_models,
-            {},
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        qrcg._coordinate_to_measure = Coordinate([1, 2, 4, 5])
-
-        rc = qrcg._get_next_run_config()
-
-        self.assertEqual(len(rc.model_run_configs()), 2)
-        mc1 = rc.model_run_configs()[0].model_config()
-        pc1 = rc.model_run_configs()[0].perf_config()
-        mc2 = rc.model_run_configs()[1].model_config()
-        pc2 = rc.model_run_configs()[1].perf_config()
-
-        self.assertEqual(mc1.to_dict(), expected_model_config1)
-        self.assertEqual(mc2.to_dict(), expected_model_config2)
-        self.assertEqual(pc1["concurrency-range"], 12)
-        self.assertEqual(pc1["batch-size"], 1)
-        self.assertEqual(pc1["model-version"], 2)
-        self.assertEqual(pc2["concurrency-range"], 192)
-        self.assertEqual(pc2["batch-size"], 1)
-        self.assertEqual(pc2["model-version"], 3)
-
-    def test_default_config_generation(self):
-        """
-        Test that the default config is generated correctly
-        """
-
-        fake_config = {
-            "name": "my-model",
-            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-            "max_batch_size": 4,
-        }
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "/tmp",
-            "--config-file",
-            "/tmp/my_config.yml",
-        ]
-
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                - my-model:
-                    perf_analyzer_flags:
-                        percentile: 96
-            """)
-        # yapf: enable
-
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            return_value=fake_config,
-        ):
-            models = [
-                ModelProfileSpec(
-                    spec=config.profile_models[0],
-                    config=config,
-                    client=MagicMock(),
-                    gpus=MagicMock(),
-                )
-            ]
-
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension(
-                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-                SearchDimension(
-                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
-                ),
-                SearchDimension(
-                    "concurrency", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-            ],
-        )
-
-        sc = SearchConfig(dimensions=dims, radius=5, min_initialized=2)
-        qrcg = QuickRunConfigGenerator(
-            sc, config, ["GPU0"], models, {}, MagicMock(), ModelVariantNameManager()
-        )
-
-        default_run_config = qrcg._create_default_run_config()
-
-        self.assertIn("--percentile=96", default_run_config.representation())
-        self.assertIn(
-            "--concurrency-range=8",
-            default_run_config.model_run_configs()[0].perf_config().representation(),
-        )
-
-    def test_default_ensemble_config_generation(self):
-        """
-        Test that the default ensemble config is generated correctly
-        """
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "/tmp",
-            "--config-file",
-            "/tmp/my_config.yml",
-        ]
-
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                - my-model:
-                    perf_analyzer_flags:
-                        percentile: 96
-            """)
-        # yapf: enable
-
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            side_effect=mock_ensemble_configs,
-        ):
-            models = [
-                ModelProfileSpec(
-                    spec=config.profile_models[0],
-                    config=config,
-                    client=MagicMock(),
-                    gpus=MagicMock(),
-                )
-            ]
-
-        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            side_effect=mock_ensemble_configs,
-        ):
-            ensemble_composing_models = (
-                RunConfigGeneratorFactory._create_composing_models(
-                    models, config, MagicMock(), MagicMock()
-                )
-            )
-
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            models,
-            ensemble_composing_models,
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        default_run_config = qrcg._create_default_run_config()
-        ensemble_composing_config_variants = default_run_config.model_run_configs()[
-            0
-        ].composing_config_variants()
-
-        self.assertIn("my-model", default_run_config.representation())
-        self.assertEqual(
-            ensemble_composing_config_variants[0].variant_name,
-            "fake_model_A_config_default",
-        )
-        self.assertEqual(
-            ensemble_composing_config_variants[1].variant_name,
-            "fake_model_B_config_default",
-        )
-
-    def test_default_bls_config_generation(self):
-        """
-        Test that the default BLS config is generated correctly
-        """
-
-        fake_config = {
-            "name": "my-model",
-            "platform": "pytorch",
-            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": [16]}],
-            "max_batch_size": 4,
-        }
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "/tmp",
-            "--config-file",
-            "/tmp/my_config.yml",
-            "--bls-composing-models",
-            "bls_composing_modelA,bls_composing_modelB",
-        ]
-
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                - my-model:
-                    perf_analyzer_flags:
-                        percentile: 96
-            """)
-        # yapf: enable
-
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            return_value=fake_config,
-        ):
-            models = [
-                ModelProfileSpec(
-                    spec=config.profile_models[0],
-                    config=config,
-                    client=MagicMock(),
-                    gpus=MagicMock(),
-                )
-            ]
-
-        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            return_value=fake_config,
-        ):
-            bls_composing_models = RunConfigGeneratorFactory._create_composing_models(
-                models, config, MagicMock(), MagicMock()
-            )
-
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            models,
-            bls_composing_models,
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        default_run_config = qrcg._create_default_run_config()
-        bls_composing_config_variants = default_run_config.model_run_configs()[
-            0
-        ].composing_config_variants()
-
-        self.assertIn("my-model", default_run_config.representation())
-        self.assertEqual(
-            bls_composing_config_variants[0].variant_name,
-            "bls_composing_modelA_config_default",
-        )
-        self.assertEqual(
-            bls_composing_config_variants[1].variant_name,
-            "bls_composing_modelB_config_default",
-        )
-
-    def test_get_next_run_config_ensemble(self):
-        """
-        Test that get_next_run_config() creates a proper RunConfig for ensemble
-        """
-        self._get_next_run_config_ensemble()
-
-    def test_get_next_run_config_ensemble_with_max_concurrency(self):
-        """
-        Test that get_next_run_config() creates a proper RunConfig for ensemble with a max concurrency
-        """
-        self._get_next_run_config_ensemble(max_concurrency=8)
-
-    def test_get_next_run_config_ensemble_with_min_concurrency(self):
-        """
-        Test that get_next_run_config() creates a proper RunConfig for ensemble with a min concurrency
-        """
-        self._get_next_run_config_ensemble(min_concurrency=16)
-
-    def test_get_next_run_config_bls(self):
-        """
-        Test that get_next_run_config() creates a proper RunConfig for BLS
-        """
-        self._get_next_run_config_bls()
-
-    def test_get_next_run_config_bls_with_max_concurrency(self):
-        """
-        Test that get_next_run_config() creates a proper RunConfig for BLS with a max concurrency
-        """
-        self._get_next_run_config_bls(max_concurrency=8)
-
-    def test_get_next_run_config_bls_with_min_concurrency(self):
-        """
-        Test that get_next_run_config() creates a proper RunConfig for BLS with a min concurrency
-        """
-        self._get_next_run_config_bls(min_concurrency=16)
-
-    def test_get_next_run_config_max_batch_size(self):
-        """
-        Test that run-config-search-max-model-batch-size is enforced
-
-        Sets up a case where the coordinate is [5,7], which corresponds to
-          - max_batch_size = 32 (will be capped at 16)
-          - instance_count = 8
-          - concurrency = 32*8*2 = 512 (will now be 16*8*2 = 256)
-
-        Also
-        - dynamic batching should be on
-        - existing values from the base model config should persist if they aren't overwritten
-        """
-        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
-        config = self._create_config(
-            additional_args=["--run-config-search-max-model-batch-size", "16"]
-        )
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            self._mock_models,
-            {},
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        qrcg._coordinate_to_measure = Coordinate([5, 7])
-
-        expected_model_config = {
-            "dynamicBatching": {},
-            "instanceGroup": [
-                {
-                    "count": 8,
-                    "kind": "KIND_GPU",
-                }
-            ],
-            "maxBatchSize": 16,
-            "name": "fake_model_name",
-            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
-        }
-        # yapf: enable
-
-        rc = qrcg._get_next_run_config()
-
-        self.assertEqual(len(rc.model_run_configs()), 1)
-        model_config = rc.model_run_configs()[0].model_config()
-        perf_config = rc.model_run_configs()[0].perf_config()
-
-        self.assertEqual(model_config.to_dict(), expected_model_config)
-        self.assertEqual(perf_config["concurrency-range"], 256)
-        self.assertEqual(perf_config["batch-size"], 1)
-
-    def test_get_next_run_config_max_instance_count(self):
-        """
-        Test that run-config-search-max-instance-count is enforced
-
-        Sets up a case where the coordinate is [5,7], which corresponds to
-          - max_batch_size = 32
-          - instance_count = 8 (will be capped at 4)
-          - concurrency = 32*8*2 = 512 (will now be 32*4*2 = 256)
-
-        Also
-        - dynamic batching should be on
-        - existing values from the base model config should persist if they aren't overwritten
-        """
-        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
-        config = self._create_config(
-            additional_args=["--run-config-search-max-instance-count", "4"]
-        )
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            self._mock_models,
-            {},
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        qrcg._coordinate_to_measure = Coordinate([5, 7])
-
-        expected_model_config = {
-            "dynamicBatching": {},
-            "instanceGroup": [
-                {
-                    "count": 4,
-                    "kind": "KIND_GPU",
-                }
-            ],
-            "maxBatchSize": 32,
-            "name": "fake_model_name",
-            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
-        }
-        # yapf: enable
-
-        rc = qrcg._get_next_run_config()
-
-        self.assertEqual(len(rc.model_run_configs()), 1)
-        model_config = rc.model_run_configs()[0].model_config()
-        perf_config = rc.model_run_configs()[0].perf_config()
-
-        self.assertEqual(model_config.to_dict(), expected_model_config)
-        self.assertEqual(perf_config["concurrency-range"], 256)
-        self.assertEqual(perf_config["batch-size"], 1)
-
-    def test_get_next_run_config_min_batch_size(self):
-        """
-        Test that run-config-search-min-model-batch-size is enforced
-
-        Sets up a case where the coordinate is [5,7], which corresponds to
-          - max_batch_size = 32 (will be min of 64)
-          - instance_count = 8
-          - concurrency = 32*8*2 = 512 (will now be 64*8*2 = 1024)
-
-        Also
-        - dynamic batching should be on
-        - existing values from the base model config should persist if they aren't overwritten
-        """
-        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
-        config = self._create_config(
-            additional_args=["--run-config-search-min-model-batch-size", "64"]
-        )
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            self._mock_models,
-            {},
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        qrcg._coordinate_to_measure = Coordinate([5, 7])
-
-        expected_model_config = {
-            "dynamicBatching": {},
-            "instanceGroup": [
-                {
-                    "count": 8,
-                    "kind": "KIND_GPU",
-                }
-            ],
-            "maxBatchSize": 64,
-            "name": "fake_model_name",
-            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
-        }
-        # yapf: enable
-
-        rc = qrcg._get_next_run_config()
-
-        self.assertEqual(len(rc.model_run_configs()), 1)
-        model_config = rc.model_run_configs()[0].model_config()
-        perf_config = rc.model_run_configs()[0].perf_config()
-
-        self.assertEqual(model_config.to_dict(), expected_model_config)
-        self.assertEqual(perf_config["concurrency-range"], 1024)
-        self.assertEqual(perf_config["batch-size"], 1)
-
-    def test_get_next_run_config_min_instance_count(self):
-        """
-        Test that run-config-search-min-instance-count is enforced
-
-        Sets up a case where the coordinate is [5,7], which corresponds to
-          - max_batch_size = 32
-          - instance_count = 8 (will be min of 16)
-          - concurrency = 32*8*2 = 512 (will now be 32*16*2 = 1024)
-
-        Also
-        - dynamic batching should be on
-        - existing values from the base model config should persist if they aren't overwritten
-        """
-        sc = SearchConfig(dimensions=self._dims, radius=5, min_initialized=2)
-        config = self._create_config(
-            additional_args=["--run-config-search-min-instance-count", "16"]
-        )
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            self._mock_models,
-            {},
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        qrcg._coordinate_to_measure = Coordinate([5, 7])
-
-        expected_model_config = {
-            "dynamicBatching": {},
-            "instanceGroup": [
-                {
-                    "count": 16,
-                    "kind": "KIND_GPU",
-                }
-            ],
-            "maxBatchSize": 32,
-            "name": "fake_model_name",
-            "input": [{"name": "INPUT__0", "dataType": "TYPE_FP32", "dims": ["16"]}],
-        }
-        # yapf: enable
-
-        rc = qrcg._get_next_run_config()
-
-        self.assertEqual(len(rc.model_run_configs()), 1)
-        model_config = rc.model_run_configs()[0].model_config()
-        perf_config = rc.model_run_configs()[0].perf_config()
-
-        self.assertEqual(model_config.to_dict(), expected_model_config)
-        self.assertEqual(perf_config["concurrency-range"], 1024)
-        self.assertEqual(perf_config["batch-size"], 1)
-
-    def test_ensemble_in_composing_models(self):
-        """
-        Test that an ensemble as a composing model raises an exception
-        """
-        additional_args = []
-        config = self._create_config(additional_args)
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            side_effect=mock_ensemble_configs,
-        ):
-            models = [
-                ModelProfileSpec(
-                    spec=config.profile_models[0],
-                    config=config,
-                    client=MagicMock(),
-                    gpus=MagicMock(),
-                )
-            ]
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            side_effect=mock_composing_ensemble_configs,
-        ) and self.assertRaises(TritonModelAnalyzerException):
-            RunConfigGeneratorFactory._create_composing_models(
-                models, config, MagicMock(), MagicMock()
-            )
-
-    def _get_next_run_config_ensemble(self, max_concurrency=0, min_concurrency=0):
-        """
-        Test that get_next_run_config() creates a proper RunConfig for ensemble
-
-        Sets up a case where the coordinate is [1,2,4,5], which corresponds to
-          - composing model A max_batch_size = 2
-          - composing model A instance_count = 3
-          - composing model A concurrency = 2*3*2 = 12
-          - composing model B max_batch_size = 16
-          - composing model B instance_count = 6
-          - composing model B concurrency = 16*6*2 = 192
-          - ensemble model concurrency = 12 (minimum value of [12, 192])
-
-        Also,
-        - sequence batching should be on for model A
-        - dynamic batching should be on for model B
-        - existing values from the base model config should persist if they aren't overwritten
-        - existing values for perf-analyzer config should persist if they aren't overwritten
-        """
-
-        additional_args = ["--cpu-only-composing-models", "fake_model_B"]
-        if max_concurrency:
-            additional_args.append("--run-config-search-max-concurrency")
-            additional_args.append(f"{max_concurrency}")
-        if min_concurrency:
-            additional_args.append("--run-config-search-min-concurrency")
-            additional_args.append(f"{min_concurrency}")
-
-        # yapf: disable
-        expected_model_A_config_0 = {
-            'instanceGroup': [{
-                'count': 3,
-                'kind': 'KIND_GPU',
-            }],
-            'maxBatchSize': 2,
-            'sequenceBatching': {},
-            'name': 'fake_model_A',
-            'input': [{
-                "name": "INPUT__0",
-                "dataType": "TYPE_FP32",
-                "dims": ['16']
-            }]
-        }
-
-        expected_model_B_config_0 = {
-            'dynamicBatching': {},
-            'instanceGroup': [{
-                'count': 6,
-                'kind': 'KIND_CPU',
-            }],
-            'maxBatchSize': 16,
-            'name': 'fake_model_B',
-            'input': [{
-                "name": "INPUT__2",
-                "dataType": "TYPE_FP16",
-                "dims": ['32']
-            }]
-        }
-        # yapf: enable
-
-        config = self._create_config(additional_args)
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            side_effect=mock_ensemble_configs,
-        ):
-            models = [
-                ModelProfileSpec(
-                    spec=config.profile_models[0],
-                    config=config,
-                    client=MagicMock(),
-                    gpus=MagicMock(),
-                )
-            ]
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            side_effect=mock_ensemble_configs,
-        ):
-            ensemble_composing_models = (
-                RunConfigGeneratorFactory._create_composing_models(
-                    models, config, MagicMock(), MagicMock()
-                )
-            )
-
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension(
-                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-                SearchDimension(
-                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
-                ),
-            ],
-        )
-        dims.add_dimensions(
-            1,
-            [
-                SearchDimension(
-                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-                SearchDimension(
-                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
-                ),
-            ],
-        )
-
-        sc = SearchConfig(dimensions=dims, radius=5, min_initialized=2)
-
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            models,
-            ensemble_composing_models,
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        qrcg._coordinate_to_measure = Coordinate([1, 2, 4, 5])
-
-        run_config = qrcg._get_next_run_config()
-
-        self.assertEqual(len(run_config.model_run_configs()), 1)
-        self.assertEqual(
-            len(run_config.model_run_configs()[0].composing_config_variants()), 2
-        )
-
-        model_config = run_config.model_run_configs()[0].model_config()
-        perf_config = run_config.model_run_configs()[0].perf_config()
-        composing_model_A_config_0 = (
-            run_config.model_run_configs()[0]
-            .composing_config_variants()[0]
-            .model_config
-        )
-        composing_model_B_config_0 = (
-            run_config.model_run_configs()[0]
-            .composing_config_variants()[1]
-            .model_config
-        )
-
-        self.assertIsInstance(model_config, ModelConfig)
-        self.assertEqual(
-            composing_model_A_config_0.to_dict(), expected_model_A_config_0
-        )
-        self.assertEqual(
-            composing_model_B_config_0.to_dict(), expected_model_B_config_0
-        )
-
-        if max_concurrency:
-            self.assertEqual(perf_config["concurrency-range"], max_concurrency)
-        elif min_concurrency:
-            self.assertEqual(perf_config["concurrency-range"], min_concurrency)
-        else:
-            self.assertEqual(perf_config["concurrency-range"], 12)
-
-        self.assertEqual(perf_config["batch-size"], 1)
-
-    def _get_next_run_config_bls(self, max_concurrency=0, min_concurrency=0):
-        """
-        Test that get_next_run_config() creates a proper RunConfig for BLS
-
-        Sets up a case where the coordinate is [1,2,3,4,5,6], which corresponds to
-          - model max_batch_size = 2
-          - model instance count = 3
-          - composing model 1 max_batch_size = 8
-          - composing model 1 instance_count = 5
-          - composing model 2 max_batch_size = 32
-          - composing model 2 instance_count = 7
-
-        Concurrency calculations are as follows:
-          - model concurrency = 2*3*2 = 12
-          - composing model 1 concurrency = 8*5*2 = 80
-          - composing model 2 concurrency = 32*7*2 = 448
-          - model concurrency = 12 (minimum value of [12, 80, 448])
-
-        Also,
-        - sequence batching should be on for model 1
-        - dynamic batching should be on for model 2
-        - existing values from the base model config should persist if they aren't overwritten
-        - existing values for perf-analyzer config should persist if they aren't overwritten
-        """
-
-        additional_args = ["--bls-composing-models", "fake_model_A,fake_model_B"]
-        if max_concurrency:
-            additional_args.append("--run-config-search-max-concurrency")
-            additional_args.append(f"{max_concurrency}")
-        if min_concurrency:
-            additional_args.append("--run-config-search-min-concurrency")
-            additional_args.append(f"{min_concurrency}")
-
-        # yapf: disable
-        expected_model_config = {
-            'instanceGroup': [{
-                'count': 3,
-                'kind': 'KIND_GPU',
-            }],
-            'maxBatchSize': 2,
-            'dynamicBatching': {},
-            'name': 'my-model',
-            'platform': 'bls',
-            'input': [{
-                "name": "INPUT__0",
-                "dataType": "TYPE_FP32",
-                "dims": ['16']
-            }]
-        }
-
-        expected_composing_model_config0 = {
-            'instanceGroup': [{
-                'count': 5,
-                'kind': 'KIND_GPU',
-            }],
-            'maxBatchSize': 8,
-            'sequenceBatching': {},
-            'name': 'fake_model_A',
-            'input': [{
-                "name": "INPUT__0",
-                "dataType": "TYPE_FP32",
-                "dims": ['16']
-            }]
-        }
-
-        expected_composing_model_config1 = {
-            'dynamicBatching': {},
-            'instanceGroup': [{
-                'count': 7,
-                'kind': 'KIND_GPU',
-            }],
-            'maxBatchSize': 32,
-            'name': 'fake_model_B',
-            'input': [{
-                "name": "INPUT__2",
-                "dataType": "TYPE_FP16",
-                "dims": ['32']
-            }]
-        }
-        # yapf: enable
-
-        config = self._create_config(additional_args)
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            side_effect=mock_bls_configs,
-        ):
-            models = [
-                ModelProfileSpec(
-                    spec=config.profile_models[0],
-                    config=config,
-                    client=MagicMock(),
-                    gpus=MagicMock(),
-                )
-            ]
-
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension(
-                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-                SearchDimension(
-                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
-                ),
-            ],
-        )
-        dims.add_dimensions(
-            1,
-            [
-                SearchDimension(
-                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-                SearchDimension(
-                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
-                ),
-            ],
-        )
-        dims.add_dimensions(
-            2,
-            [
-                SearchDimension(
-                    "max_batch_size", SearchDimension.DIMENSION_TYPE_EXPONENTIAL
-                ),
-                SearchDimension(
-                    "instance_count", SearchDimension.DIMENSION_TYPE_LINEAR
-                ),
-            ],
-        )
-
-        sc = SearchConfig(dimensions=dims, radius=5, min_initialized=2)
-
-        with patch(
-            "model_analyzer.triton.model.model_config.ModelConfig.create_model_config_dict",
-            side_effect=mock_bls_configs,
-        ):
-            bls_composing_models = RunConfigGeneratorFactory._create_composing_models(
-                models, config, MagicMock(), MagicMock()
-            )
-
-        qrcg = QuickRunConfigGenerator(
-            sc,
-            config,
-            MagicMock(),
-            models,
-            bls_composing_models,
-            MagicMock(),
-            ModelVariantNameManager(),
-        )
-
-        qrcg._coordinate_to_measure = Coordinate([1, 2, 3, 4, 5, 6])
-
-        run_config = qrcg._get_next_run_config()
-
-        self.assertEqual(len(run_config.model_run_configs()), 1)
-
-        self.assertEqual(
-            len(run_config.model_run_configs()[0].composing_config_variants()), 2
-        )
-
-        model_config = run_config.model_run_configs()[0].model_config()
-        perf_config = run_config.model_run_configs()[0].perf_config()
-        composing_model_config0 = (
-            run_config.model_run_configs()[0]
-            .composing_config_variants()[0]
-            .model_config
-        )
-        composing_model_config1 = (
-            run_config.model_run_configs()[0]
-            .composing_config_variants()[1]
-            .model_config
-        )
-
-        self.assertEqual(model_config.to_dict(), expected_model_config)
-        self.assertEqual(
-            composing_model_config0.to_dict(), expected_composing_model_config0
-        )
-        self.assertEqual(
-            composing_model_config1.to_dict(), expected_composing_model_config1
-        )
-
-        if max_concurrency:
-            self.assertEqual(perf_config["concurrency-range"], max_concurrency)
-        elif min_concurrency:
-            self.assertEqual(perf_config["concurrency-range"], min_concurrency)
-        else:
-            self.assertEqual(perf_config["concurrency-range"], 12)
-
-        self.assertEqual(perf_config["batch-size"], 1)
-
-    def _create_config(self, additional_args=[]):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "/tmp",
-            "--config-file",
-            "/tmp/my_config.yml",
-        ]
-
-        for arg in additional_args:
-            args.append(arg)
-
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                - my-model
-            """)
-        # yapf: enable
-
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        return config
-
-    def tearDown(self):
-        patch.stopall()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_record_aggregator.py b/src/c++/perf_analyzer/tests/test_record_aggregator.py
deleted file mode 100755
index f95165296..000000000
--- a/src/c++/perf_analyzer/tests/test_record_aggregator.py
+++ /dev/null
@@ -1,175 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.record.record_aggregator import RecordAggregator
-from model_analyzer.record.types.gpu_utilization import GPUUtilization
-from model_analyzer.record.types.perf_latency_p99 import PerfLatencyP99
-from model_analyzer.record.types.perf_throughput import PerfThroughput
-
-from .common import test_result_collector as trc
-
-
-class TestRecordAggregatorMethods(trc.TestResultCollector):
-    def tearDown(self):
-        patch.stopall()
-
-    def test_insert(self):
-        record_aggregator = RecordAggregator()
-
-        self.assertEqual(record_aggregator.total(), 0)
-
-        throughput_record = PerfThroughput(5)
-        record_aggregator.insert(throughput_record)
-
-        # Assert record is added
-        self.assertEqual(record_aggregator.total(), 1)
-
-    def test_record_types(self):
-        record_aggregator = RecordAggregator()
-
-        throughput_record = PerfThroughput(5)
-        record_aggregator.insert(throughput_record)
-
-        self.assertEqual(record_aggregator.record_types()[0], PerfThroughput)
-
-    def test_filter_records_default(self):
-        record_aggregator = RecordAggregator()
-
-        # insert throughput record and check its presence
-        throughput_record = PerfThroughput(5)
-        record_aggregator.insert(throughput_record)
-
-        # Get the record
-        retrieved_records = record_aggregator.filter_records().get_records()
-        retrieved_throughput = retrieved_records[PerfThroughput][0]
-
-        self.assertIsInstance(
-            retrieved_throughput,
-            PerfThroughput,
-            msg="Record types do not match after filter_records",
-        )
-
-        self.assertEqual(
-            retrieved_throughput.value(),
-            throughput_record.value(),
-            msg="Values do not match after filter_records",
-        )
-
-    def test_filter_records_filtered(self):
-        record_aggregator = RecordAggregator()
-
-        # Test for malformed inputs
-        with self.assertRaises(Exception):
-            record_aggregator.filter_records(filters=[(lambda x: False)])
-        with self.assertRaises(Exception):
-            record_aggregator.filter_records(
-                record_types=[None, None], filters=[(lambda x: False)]
-            )
-
-        # Insert 3 throughputs
-        record_aggregator.insert(PerfThroughput(5))
-        record_aggregator.insert(PerfThroughput(1))
-        record_aggregator.insert(PerfThroughput(10))
-
-        # Test get with filters
-        retrieved_records = record_aggregator.filter_records(
-            record_types=[PerfThroughput], filters=[(lambda v: v.value() >= 5)]
-        ).get_records()
-
-        # Should return 2 records
-        self.assertEqual(len(retrieved_records[PerfThroughput]), 2)
-        retrieved_values = [
-            record.value() for record in retrieved_records[PerfThroughput]
-        ]
-        self.assertIn(5, retrieved_values)
-        self.assertIn(10, retrieved_values)
-
-        # Insert 2 Latency records
-        record_aggregator.insert(PerfLatencyP99(3))
-        record_aggregator.insert(PerfLatencyP99(6))
-
-        # Test get with multiple headers
-        retrieved_records = record_aggregator.filter_records(
-            record_types=[PerfLatencyP99, PerfThroughput],
-            filters=[(lambda v: v.value() == 3), (lambda v: v.value() < 5)],
-        ).get_records()
-
-        retrieved_values = {
-            record_type: [record.value() for record in retrieved_records[record_type]]
-            for record_type in [PerfLatencyP99, PerfThroughput]
-        }
-
-        self.assertEqual(len(retrieved_records[PerfLatencyP99]), 1)
-        self.assertIn(3, retrieved_values[PerfLatencyP99])
-
-        self.assertEqual(len(retrieved_records[PerfThroughput]), 1)
-        self.assertIn(1, retrieved_values[PerfThroughput])
-
-    def test_groupby(self):
-        record_aggregator = RecordAggregator()
-        # Insert 3 throughputs
-        record_aggregator.insert(PerfThroughput(5, timestamp=0))
-        record_aggregator.insert(PerfThroughput(1, timestamp=1))
-        record_aggregator.insert(PerfThroughput(10, timestamp=1))
-
-        def groupby_criteria(record):
-            return record.timestamp()
-
-        records = record_aggregator.groupby([PerfThroughput], groupby_criteria)
-        self.assertEqual(list(records[PerfThroughput]), [0, 1])
-        self.assertEqual(
-            list(records[PerfThroughput].values()),
-            [PerfThroughput(5.0), PerfThroughput(10.0)],
-        )
-
-        records = record_aggregator.groupby([PerfThroughput], groupby_criteria)
-        self.assertEqual(list(records[PerfThroughput]), [0, 1])
-        self.assertEqual(
-            list(records[PerfThroughput].values()),
-            [PerfThroughput(5.0), PerfThroughput(10.0)],
-        )
-
-    def test_aggregate(self):
-        record_aggregator = RecordAggregator()
-
-        # Insert 10 records
-        for i in range(10):
-            record_aggregator.insert(PerfThroughput(i))
-
-        for i in range(10):
-            record_aggregator.insert(GPUUtilization(i))
-        # Aggregate them with max, min and average
-        max_vals = record_aggregator.aggregate(record_types=[PerfThroughput])
-        avg_vals = record_aggregator.aggregate(record_types=[GPUUtilization])
-
-        self.assertEqual(
-            max_vals[PerfThroughput],
-            PerfThroughput(9),
-            msg="Aggregation failed with max",
-        )
-
-        self.assertEqual(
-            avg_vals[GPUUtilization],
-            GPUUtilization(4.5),
-            msg="Aggregation failed with max",
-        )
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_record_types.py b/src/c++/perf_analyzer/tests/test_record_types.py
deleted file mode 100755
index 4bd6d8b32..000000000
--- a/src/c++/perf_analyzer/tests/test_record_types.py
+++ /dev/null
@@ -1,191 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.record.record import RecordType
-
-from .common import test_result_collector as trc
-
-
-class TestRecordAggregatorMethods(trc.TestResultCollector):
-    """
-    The record types in the model-analyzer.record.types package are contextual
-        when it uses 'less than' (<) and 'greater than' (>) operators.
-
-    The 'less than' and 'greater than' operators are overloaded to
-        mean 'worse than' and 'better than' respectively.
-
-    Some record types treat MORE as better
-        (eg, gpu_free_memory, cpu_available_ram)
-    Other record types treat LESS as better
-        (eg, gpu_used_memory, cpu_used_ram)
-
-    So, when comparing two objects of type 'cpu_used_ram'
-        12 > 13 is actually true, since 12 'is better than' 13.
-    """
-
-    def setUp(self):
-        record_types = RecordType.get_all_record_types()
-        self.all_record_types = record_types.values()
-        self.less_is_better_types = {
-            record_types[k]
-            for k in [
-                "perf_latency_avg",
-                "perf_latency_p90",
-                "perf_latency_p95",
-                "perf_latency_p99",
-                "gpu_used_memory",
-                "cpu_used_ram",
-                "perf_server_compute_infer",
-                "perf_latency",
-                "perf_server_queue",
-                "perf_client_response_wait",
-                "perf_server_compute_output",
-                "perf_client_send_recv",
-                "perf_server_compute_input",
-                "gpu_power_usage",
-            ]
-        }
-        self.more_is_better_types = {
-            record_types[k]
-            for k in [
-                "perf_throughput",
-                "gpu_free_memory",
-                "gpu_utilization",
-                "cpu_available_ram",
-                "gpu_total_memory",
-            ]
-        }
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_counts(self):
-        """
-        Make sure that all 'worse than' and 'better than' tests are tested
-        """
-        total_count = len(self.all_record_types)
-        less_is_better_count = len(self.less_is_better_types)
-        more_is_better_count = len(self.more_is_better_types)
-        self.assertEqual(total_count, less_is_better_count + more_is_better_count)
-
-    def test_add(self):
-        """
-        Test __add__ function for
-        each record type
-        """
-
-        for record_type in self.all_record_types:
-            metric1 = record_type(value=5)
-            metric2 = record_type(value=9)
-            metric3 = metric1 + metric2
-            self.assertIsInstance(metric3, record_type)
-            self.assertEqual(metric3.value(), 14)
-
-    def test_sub(self):
-        """
-        Test __sub__ function for
-        each record type
-        """
-
-        for record_type in self.all_record_types:
-            metric1 = record_type(value=10)
-            metric2 = record_type(value=3)
-            metric3 = metric1 - metric2
-            self.assertIsInstance(metric3, record_type)
-            if record_type in self.less_is_better_types:
-                self.assertEqual(metric3.value(), -7)
-            elif record_type in self.more_is_better_types:
-                self.assertEqual(metric3.value(), 7)
-
-    def test_mult(self):
-        """
-        Test __mult__ function for
-        each record type
-        """
-
-        for record_type in self.all_record_types:
-            metric1 = record_type(value=6)
-            metric2 = metric1 * 2
-            self.assertIsInstance(metric2, record_type)
-            self.assertEqual(metric2.value(), 12)
-
-    def test_div(self):
-        """
-        Test __div__ function for
-        each record type
-        """
-
-        for record_type in self.all_record_types:
-            metric1 = record_type(value=60)
-            metric2 = metric1 / 12
-            self.assertIsInstance(metric2, record_type)
-            self.assertEqual(metric2.value(), 5)
-
-    def test_compare(self):
-        """
-        Test __lt__, __eq__, __gt__
-        functions for each record type
-        """
-
-        for record_type in self.all_record_types:
-            metric1 = record_type(value=10.6)
-            metric2 = record_type(value=3.2)
-
-            # Test __lt__ (True if 1 worse than 2)
-            if record_type in self.less_is_better_types:
-                self.assertLess(metric1, metric2)
-            elif record_type in self.more_is_better_types:
-                self.assertLess(metric2, metric1)
-
-            # Test __gt__ (True if 1 better than 2)
-            if record_type in self.less_is_better_types:
-                self.assertLess(metric1, metric2)
-            elif record_type in self.more_is_better_types:
-                self.assertGreater(metric1, metric2)
-
-            # Test __eq__
-            metric1 = record_type(value=12)
-            metric2 = record_type(value=12)
-            self.assertEqual(metric1, metric2)
-
-    def test_value_function(self):
-        avg_value = RecordType.get_all_record_types()[
-            "perf_latency_p99"
-        ].value_function()([10, 50, 100, 40])
-
-        total_value = RecordType.get_all_record_types()[
-            "perf_throughput"
-        ].value_function()([10, 50, 100, 40])
-
-        self.assertEqual(avg_value, 50)
-        self.assertEqual(total_value, 200)
-
-    def test_calculate_percentage_gain(self):
-        for record_type in self.all_record_types:
-            metric1 = record_type(value=10)
-            metric2 = record_type(value=5)
-
-            if record_type in self.less_is_better_types:
-                self.assertEqual(metric1.calculate_percentage_gain(metric2), -50)
-            else:
-                self.assertEqual(metric1.calculate_percentage_gain(metric2), 100)
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_remote_monitor.py b/src/c++/perf_analyzer/tests/test_remote_monitor.py
deleted file mode 100755
index 6003c3f17..000000000
--- a/src/c++/perf_analyzer/tests/test_remote_monitor.py
+++ /dev/null
@@ -1,174 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import time
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.monitor.remote_monitor import RemoteMonitor
-from model_analyzer.record.types.gpu_free_memory import GPUFreeMemory
-from model_analyzer.record.types.gpu_power_usage import GPUPowerUsage
-from model_analyzer.record.types.gpu_used_memory import GPUUsedMemory
-from model_analyzer.record.types.gpu_utilization import GPUUtilization
-from tests.mocks.mock_requests import MockRequests
-
-from .common import test_result_collector as trc
-
-TEST_DEVICE_NAME = "TEST_DEVICE_NAME"
-TEST_DEVICE_ID = 0
-TEST_METRICS_URL = "localhost:8002"
-
-TEST_POWER_USAGE = 20
-TEST_GPU_MEMORY_USAGE = 400000000
-TEST_GPU_UTILIZATION = 0.5
-TEST_TOTAL_BYTES = 1000000000
-
-TEST_METRICS_RESPONSE = bytes(
-    "# HELP nv_inference_request_success Number of successful"
-    "inference requests, all batch sizes\n# TYPE nv_inference_request_success counter\n# "
-    "HELP nv_inference_request_failure Number of failed inference requests, all batch sizes\n# "
-    "TYPE nv_inference_request_failure counter\n# HELP nv_inference_count Number of inferences "
-    "performed\n# TYPE nv_inference_count counter\n# HELP nv_inference_exec_count Number of model"
-    "executions performed\n# TYPE nv_inference_exec_count counter\n# "
-    "HELP nv_inference_request_duration_us Cumulative inference request duration in microseconds\n# "
-    "TYPE nv_inference_request_duration_us counter\n# HELP nv_inference_queue_duration_us Cumulative "
-    "inference queuing duration in microseconds\n# TYPE nv_inference_queue_duration_us counter\n# HELP "
-    "nv_inference_compute_input_duration_us Cumulative compute input duration in microseconds\n# TYPE "
-    "nv_inference_compute_input_duration_us counter\n# HELP nv_inference_compute_infer_duration_us Cumulative "
-    "compute inference duration in microseconds\n# TYPE nv_inference_compute_infer_duration_us counter\n# HELP "
-    "nv_inference_compute_output_duration_us Cumulative inference compute output duration in microseconds\n# TYPE "
-    "nv_inference_compute_output_duration_us counter\n# HELP nv_gpu_utilization GPU utilization rate [0.0 - 1.0)\n# "
-    'TYPE nv_gpu_utilization gauge\nnv_gpu_utilization{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} '
-    f"{TEST_GPU_UTILIZATION}\n# HELP nv_gpu_memory_total_bytes GPU total memory, in bytes\n# TYPE nv_gpu_memory_total_bytes "
-    'gauge\nnv_gpu_memory_total_bytes{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} '
-    f"{TEST_TOTAL_BYTES}\n# HELP nv_gpu_memory_used_bytes GPU used memory, in bytes\n# TYPE "
-    'nv_gpu_memory_used_bytes gauge\nnv_gpu_memory_used_bytes{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} '
-    f"{TEST_GPU_MEMORY_USAGE}\n# HELP nv_gpu_power_usage GPU power usage in watts\n# TYPE nv_gpu_power_usage "
-    'gauge\nnv_gpu_power_usage{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} '
-    f"{TEST_POWER_USAGE}\n# HELP nv_gpu_power_limit GPU power management limit in watts\n# TYPE "
-    'nv_gpu_power_limit gauge\nnv_gpu_power_limit{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} 280.000000\n# '
-    "HELP nv_energy_consumption GPU energy consumption in joules since the Triton Server started\n# TYPE nv_energy_consumption "
-    'counter\nnv_energy_consumption{gpu_uuid="GPU-e35ba3d2-6eef-2bb9-e35c-6ef6eada4f11"} 1474.042000\n',
-    encoding="ascii",
-)
-
-
-class TestRemoteMonitor(trc.TestResultCollector):
-    def setUp(self):
-        self.mock_requests = MockRequests(
-            mock_paths=["model_analyzer.monitor.remote_monitor"]
-        )
-        self.mock_requests.start()
-        self.mock_requests.set_get_request_response(TEST_METRICS_RESPONSE)
-
-    def test_record_memory(self):
-        # One measurement every 0.1 seconds
-        frequency = 0.1
-        monitoring_time = 0.1
-        metrics = [GPUUsedMemory, GPUFreeMemory]
-        gpu_monitor = RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
-        gpu_monitor.start_recording_metrics()
-        time.sleep(monitoring_time)
-        records = gpu_monitor.stop_recording_metrics()
-
-        # Assert instance types
-        for record in records:
-            self.assertIsInstance(record.device_uuid(), str)
-            self.assertIsInstance(record.value(), float)
-            if isinstance(record, GPUFreeMemory):
-                self.assertEqual(
-                    record.value(), (TEST_TOTAL_BYTES - TEST_GPU_MEMORY_USAGE) // 1e6
-                )
-            else:
-                self.assertEqual(record.value(), TEST_GPU_MEMORY_USAGE // 1e6)
-            self.assertIsInstance(record.timestamp(), int)
-
-        # The number of records should be dividable by number of metrics
-        self.assertEqual(len(records) % len(metrics), 0)
-        self.assertGreater(len(records), 0)
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            gpu_monitor.stop_recording_metrics()
-
-        gpu_monitor.destroy()
-
-        metrics = ["UndefinedTag"]
-        with self.assertRaises(TritonModelAnalyzerException):
-            RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
-
-    def test_record_power(self):
-        # One measurement every 0.01 seconds
-        frequency = 0.1
-        monitoring_time = 0.1
-        metrics = [GPUPowerUsage]
-        gpu_monitor = RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
-        gpu_monitor.start_recording_metrics()
-        time.sleep(monitoring_time)
-        records = gpu_monitor.stop_recording_metrics()
-
-        # Assert instance types
-        for record in records:
-            self.assertIsInstance(record.device_uuid(), str)
-            self.assertIsInstance(record.value(), float)
-            self.assertEqual(record.value(), TEST_POWER_USAGE)
-            self.assertIsInstance(record.timestamp(), int)
-
-        # The number of records should be dividable by number of metrics
-        self.assertEqual(len(records) % len(metrics), 0)
-        self.assertGreater(len(records), 0)
-
-        gpu_monitor.destroy()
-
-    def test_record_utilization(self):
-        # One measurement every 0.01 seconds
-        frequency = 0.1
-        monitoring_time = 0.1
-        metrics = [GPUUtilization]
-        gpu_monitor = RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
-        gpu_monitor.start_recording_metrics()
-        time.sleep(monitoring_time)
-        records = gpu_monitor.stop_recording_metrics()
-
-        # Assert instance types
-        for record in records:
-            self.assertIsInstance(record.device_uuid(), str)
-            self.assertIsInstance(record.value(), float)
-            self.assertLessEqual(record.value(), 100)
-            self.assertEqual(record.value(), TEST_GPU_UTILIZATION * 100)
-            self.assertIsInstance(record.timestamp(), int)
-
-        # The number of records should be dividable by number of metrics
-        self.assertEqual(len(records) % len(metrics), 0)
-        self.assertGreater(len(records), 0)
-
-        gpu_monitor.destroy()
-
-    def test_immediate_start_stop(self):
-        frequency = 1
-        metrics = [GPUUsedMemory, GPUFreeMemory]
-        gpu_monitor = RemoteMonitor(TEST_METRICS_URL, frequency, metrics)
-        gpu_monitor.start_recording_metrics()
-        gpu_monitor.stop_recording_metrics()
-        gpu_monitor.destroy()
-
-    def tearDown(self):
-        patch.stopall()
-        self.mock_requests.stop()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_report_manager.py b/src/c++/perf_analyzer/tests/test_report_manager.py
deleted file mode 100755
index 6a4f26616..000000000
--- a/src/c++/perf_analyzer/tests/test_report_manager.py
+++ /dev/null
@@ -1,642 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.config.run.model_run_config import ModelRunConfig
-from model_analyzer.config.run.run_config import RunConfig
-from model_analyzer.perf_analyzer.perf_config import PerfAnalyzerConfig
-from model_analyzer.reports.report_manager import ReportManager
-from model_analyzer.result.constraint_manager import ConstraintManager
-from model_analyzer.result.result_manager import ResultManager
-from model_analyzer.result.run_config_result_comparator import RunConfigResultComparator
-from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
-from model_analyzer.triton.model.model_config import ModelConfig
-from model_analyzer.triton.model.model_config_variant import ModelConfigVariant
-
-from .common import test_result_collector as trc
-from .common.test_utils import (
-    construct_constraint_manager,
-    construct_run_config_measurement,
-    evaluate_mock_config,
-)
-from .mocks.mock_io import MockIOMethods
-from .mocks.mock_json import MockJSONMethods
-from .mocks.mock_matplotlib import MockMatplotlibMethods
-from .mocks.mock_os import MockOSMethods
-
-
-class TestReportManager(trc.TestResultCollector):
-    def _init_managers(
-        self,
-        models="test_model",
-        num_configs_per_model=10,
-        mode="online",
-        subcommand="profile",
-        report_gpu_metrics=False,
-    ):
-        args = ["model-analyzer", subcommand, "-f", "path-to-config-file"]
-        if subcommand == "profile":
-            args.extend(["--profile-models", models])
-            args.extend(["--model-repository", "/tmp"])
-        else:
-            args.extend(["--report-model-configs", models])
-
-        if report_gpu_metrics:
-            args.extend(["--always-report-gpu-metrics"])
-
-        yaml_str = (
-            """
-            num_configs_per_model: """
-            + str(num_configs_per_model)
-            + """
-            client_protocol: grpc
-            export_path: /test/export/path
-            constraints:
-              perf_latency_p99:
-                max: 100
-              gpu_used_memory:
-                max: 10000
-        """
-        )
-        config = evaluate_mock_config(args, yaml_str, subcommand=subcommand)
-        state_manager = AnalyzerStateManager(config=config, server=None)
-
-        gpu_info = {"gpu_uuid": {"name": "fake_gpu_name", "total_memory": 1024000000}}
-        constraint_manager = ConstraintManager(config=config)
-
-        self.result_manager = ResultManager(
-            config=config,
-            state_manager=state_manager,
-            constraint_manager=constraint_manager,
-        )
-        self.report_manager = ReportManager(
-            mode=mode,
-            config=config,
-            gpu_info=gpu_info,
-            result_manager=self.result_manager,
-            constraint_manager=constraint_manager,
-        )
-
-    def _add_result_measurement(
-        self,
-        model_config_name,
-        model_name,
-        avg_gpu_metrics,
-        avg_non_gpu_metrics,
-        result_comparator,
-        cpu_only=False,
-        add_to_results_only=False,
-    ):
-        config_pb = self.model_config.copy()
-        config_pb["name"] = model_config_name
-        model_config = ModelConfig.create_from_dictionary(config_pb)
-        model_config_variant = ModelConfigVariant(
-            model_config, model_config_name, cpu_only
-        )
-
-        measurement = construct_run_config_measurement(
-            model_name=model_name,
-            model_config_names=[model_config_name],
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values=avg_gpu_metrics,
-            non_gpu_metric_values=[avg_non_gpu_metrics],
-            metric_objectives=result_comparator._metric_weights,
-            model_config_weights=[1],
-        )
-
-        perf_config = PerfAnalyzerConfig()
-        perf_config.update_config({"model-name": model_config_name})
-        mrc = ModelRunConfig(model_name, model_config_variant, perf_config)
-        run_config = RunConfig({})
-        run_config.add_model_run_config(mrc)
-
-        if add_to_results_only:
-            self.result_manager._add_rcm_to_results(run_config, measurement)
-        else:
-            self.result_manager.add_run_config_measurement(run_config, measurement)
-
-    def setUp(self):
-        self.model_config = {
-            "platform": "tensorflow_graphdef",
-            "instance_group": [{"count": 1, "kind": "KIND_GPU"}],
-            "max_batch_size": 8,
-            "dynamic_batching": {},
-        }
-
-        self.os_mock = MockOSMethods(
-            mock_paths=[
-                "model_analyzer.reports.report_manager",
-                "model_analyzer.plots.plot_manager",
-                "model_analyzer.state.analyzer_state_manager",
-                "model_analyzer.config.input.config_utils",
-                "model_analyzer.config.input.config_command_profile",
-                "model_analyzer.config.input.config_command_report",
-            ]
-        )
-        self.os_mock.start()
-        self.os_mock.set_os_path_isfile_return_value(False)
-        # Required patch ordering here
-        # html_report must be patched before pdf_report
-        # Likely due to patching dealing with parent + child classes
-        self.io_mock = MockIOMethods(
-            mock_paths=[
-                "model_analyzer.reports.html_report",
-                "model_analyzer.reports.pdf_report",
-                "model_analyzer.state.analyzer_state_manager",
-            ],
-            read_data=[bytes(">:(".encode("ascii"))],
-        )
-        self.io_mock.start()
-        self.matplotlib_mock = MockMatplotlibMethods()
-        self.matplotlib_mock.start()
-        self.json_mock = MockJSONMethods()
-        self.json_mock.start()
-
-    def test_add_results(self, *args):
-        for mode in ["online", "offline"]:
-            self._init_managers(
-                models="test_model1,test_model2", mode=mode, subcommand="profile"
-            )
-            result_comparator = RunConfigResultComparator(
-                metric_objectives_list=[{"perf_throughput": 10}], model_weights=[1]
-            )
-
-            avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-
-            for i in range(10):
-                avg_non_gpu_metrics = {
-                    "perf_throughput": 100 + 10 * i,
-                    "perf_latency_p99": 4000,
-                    "cpu_used_ram": 1000,
-                }
-                self._add_result_measurement(
-                    f"test_model1_config_report_{i}",
-                    "test_model1",
-                    avg_gpu_metrics,
-                    avg_non_gpu_metrics,
-                    result_comparator,
-                    add_to_results_only=False,
-                )
-
-            for i in range(5):
-                avg_non_gpu_metrics = {
-                    "perf_throughput": 200 + 10 * i,
-                    "perf_latency_p99": 4000,
-                    "cpu_used_ram": 1000,
-                }
-                self._add_result_measurement(
-                    f"test_model2_config_report_{i}",
-                    "test_model2",
-                    avg_gpu_metrics,
-                    avg_non_gpu_metrics,
-                    result_comparator,
-                    add_to_results_only=False,
-                )
-
-            self.report_manager.create_summaries()
-            self.assertEqual(
-                self.report_manager.report_keys(), ["test_model1", "test_model2"]
-            )
-
-            report1_data = self.report_manager.data("test_model1")
-            report2_data = self.report_manager.data("test_model2")
-
-            self.assertEqual(len(report1_data), 10)
-            self.assertEqual(len(report2_data), 5)
-
-    @patch(
-        "model_analyzer.reports.report_manager.ReportManager._find_default_run_config_measurement",
-        return_value=100,
-    )
-    @patch(
-        "model_analyzer.reports.report_manager.ReportManager._get_objective_gain",
-        return_value=100,
-    )
-    def test_build_summary_table(self, *args):
-        for mode in ["offline", "online"]:
-            for cpu_only in [True, False]:
-                for report_gpu_metrics in [True, False]:
-                    self.subtest_build_summary_table(mode, cpu_only, report_gpu_metrics)
-
-    def subtest_build_summary_table(self, mode, cpu_only, report_gpu_metrics):
-        self._init_managers(
-            models="test_model",
-            mode=mode,
-            subcommand="profile",
-            report_gpu_metrics=report_gpu_metrics,
-        )
-        result_comparator = RunConfigResultComparator(
-            metric_objectives_list=[{"perf_throughput": 10}], model_weights=[1]
-        )
-
-        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-
-        gpu_metrics = report_gpu_metrics or not cpu_only
-
-        for i in range(10, 0, -1):
-            avg_non_gpu_metrics = {
-                "perf_throughput": 100 + 10 * i,
-                "perf_latency_p99": 4000,
-                "cpu_used_ram": 1000,
-            }
-            self._add_result_measurement(
-                f"test_model_config_{i}",
-                "test_model",
-                avg_gpu_metrics,
-                avg_non_gpu_metrics,
-                result_comparator,
-                cpu_only=not gpu_metrics,
-            )
-
-        self.report_manager.create_summaries()
-
-        summary_table, summary_sentence = self.report_manager._build_summary_table(
-            report_key="test_model",
-            num_measurements=10,
-            num_configurations=3,
-            gpu_name="TITAN RTX",
-            report_gpu_metrics=gpu_metrics,
-        )
-
-        if mode == "online":
-            objective = "maximizing throughput"
-        else:
-            objective = "minimizing latency"
-
-        if gpu_metrics:
-            expected_summary_sentence = (
-                "In 10 measurements across 3 configurations, "
-                "<strong>test_model_config_10</strong> is <strong>100%</strong> better than the default configuration "
-                f"at {objective}, under the given constraints, on GPU(s) TITAN RTX.<UL><LI> "
-                "<strong>test_model_config_10</strong>: 1 GPU instance with a max batch size of 8 on platform tensorflow_graphdef "
-                "</LI> </UL>"
-            )
-        else:
-            expected_summary_sentence = (
-                "In 10 measurements across 3 configurations, "
-                "<strong>test_model_config_10</strong> is <strong>100%</strong> better than the default configuration "
-                f"at {objective}, under the given constraints.<UL><LI> "
-                "<strong>test_model_config_10</strong>: 1 GPU instance with a max batch size of 8 on platform tensorflow_graphdef "
-                "</LI> </UL>"
-            )
-
-        self.assertEqual(len(expected_summary_sentence), len(summary_sentence))
-        self.assertEqual(expected_summary_sentence, summary_sentence)
-
-        # Get throughput index and make sure results are sorted
-        throughput_index = summary_table.headers().index("Throughput (infer/sec)")
-        model_name_index = summary_table.headers().index("Model Config Name")
-        for i in range(9):
-            current_row = summary_table.get_row_by_index(i)
-            next_row = summary_table.get_row_by_index(i + 1)
-            self.assertEqual(current_row[model_name_index], f"test_model_config_{10-i}")
-            self.assertGreaterEqual(
-                current_row[throughput_index], next_row[throughput_index]
-            )
-
-    def test_build_detailed_info(self):
-        for cpu_only in [True, False]:
-            for report_gpu_metrics in [True, False]:
-                self._subtest_build_detailed_info(cpu_only, report_gpu_metrics)
-
-    def _subtest_build_detailed_info(self, cpu_only, report_gpu_metrics):
-        self._init_managers(models="test_model_config_10", subcommand="report")
-
-        result_comparator = RunConfigResultComparator(
-            metric_objectives_list=[{"perf_throughput": 10}], model_weights=[1]
-        )
-
-        gpu_metrics = report_gpu_metrics or not cpu_only
-
-        avg_gpu_metrics = {"gpu_uuid": {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-
-        for i in range(10, 0, -1):
-            avg_non_gpu_metrics = {
-                "perf_throughput": 100 + 10 * i,
-                "perf_latency_p99": 4000,
-                "cpu_used_ram": 1000,
-            }
-            self._add_result_measurement(
-                f"test_model_config_{i}",
-                "test_model",
-                avg_gpu_metrics,
-                avg_non_gpu_metrics,
-                result_comparator,
-                cpu_only=not gpu_metrics,
-                add_to_results_only=True,
-            )
-
-        self.report_manager._add_detailed_report_data()
-        self.report_manager._build_detailed_table("test_model_config_10")
-        sentence = self.report_manager._build_detailed_info("test_model_config_10")
-
-        if gpu_metrics:
-            expected_sentence = (
-                f"The model config <strong>test_model_config_10</strong> uses 1 GPU instance with "
-                f"a max batch size of 8 and has dynamic batching enabled. 1 measurement(s) "
-                f"were obtained for the model config on GPU(s) 1 x fake_gpu_name with total memory 1.0 GB. "
-                f"This model uses the platform tensorflow_graphdef."
-            )
-        else:
-            expected_sentence = (
-                f"The model config <strong>test_model_config_10</strong> uses 1 GPU instance with "
-                f"a max batch size of 8 and has dynamic batching enabled. 1 measurement(s) "
-                f"were obtained for the model config on CPU. "
-                f"This model uses the platform tensorflow_graphdef."
-            )
-
-        self.assertEqual(expected_sentence, sentence)
-
-    @patch("model_analyzer.plots.plot_manager.PlotManager._create_update_simple_plot")
-    @patch("model_analyzer.result.result_table.ResultTable.insert_row_by_index")
-    def test_summary_default_within_top(self, add_table_fn, add_plot_fn):
-        """
-        Test summary report generation when default is in the top n configs
-
-        Creates some results where the default config is within the top n configs,
-        and then confirms that the number of entries added to plots and tables
-        is correct
-        """
-
-        default_within_top = True
-        top_n = 3
-        self._test_summary_counts(add_table_fn, add_plot_fn, default_within_top, top_n)
-
-    @patch("model_analyzer.plots.plot_manager.PlotManager._create_update_simple_plot")
-    @patch("model_analyzer.result.result_table.ResultTable.insert_row_by_index")
-    def test_summary_default_not_within_top(self, add_table_fn, add_plot_fn):
-        """
-        Test summary report generation when default is not in the top n configs
-
-        Creates some results where the default config is not within the top n configs,
-        and then confirms that the number of entries added to plots and tables
-        is correct such that it includes the default config data
-        """
-        default_within_top = False
-        top_n = 3
-        self._test_summary_counts(add_table_fn, add_plot_fn, default_within_top, top_n)
-
-    def _test_summary_counts(
-        self, add_table_fn, add_plot_fn, default_within_top, top_n
-    ):
-        """
-        Helper function to test creating summary reports and confirming that the number
-        of entries added to plots and tables is as expected
-        """
-        num_plots_in_summary_report = 2
-        num_tables_in_summary_report = 1
-        expected_config_count = top_n + 1 if not default_within_top else top_n
-        expected_plot_count = num_plots_in_summary_report * expected_config_count
-        expected_table_count = num_tables_in_summary_report * expected_config_count
-
-        self._init_managers(
-            models="test_model1", num_configs_per_model=top_n, subcommand="profile"
-        )
-        result_comparator = RunConfigResultComparator(
-            metric_objectives_list=[{"perf_throughput": 10}], model_weights=[1]
-        )
-        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-        for i in range(10):
-            # Create a bunch of fake measurement results
-            #
-            # For the 'default_within_top' case, have the throughput
-            # be decreasing for each config (so the highest throughput is the
-            # first, default config)
-            #
-            # For the default not within top case, have the throughput be
-            # increasing for each config (so the highest throughput is the last
-            # config, and the default config is the worst)
-            #
-            p99 = 20 + i
-            throughput = 100 - 10 * i if default_within_top else 200 + 10 * i
-            avg_non_gpu_metrics = {
-                "perf_throughput": throughput,
-                "perf_latency_p99": p99,
-                "cpu_used_ram": 1000,
-            }
-            name = f"test_model1_config_{i}"
-            if not i:
-                name = f"test_model1_config_default"
-            self._add_result_measurement(
-                name,
-                "test_model1",
-                avg_gpu_metrics,
-                avg_non_gpu_metrics,
-                result_comparator,
-            )
-        self.report_manager.create_summaries()
-
-        self.assertEqual(expected_plot_count, add_plot_fn.call_count)
-        self.assertEqual(expected_table_count, add_table_fn.call_count)
-
-    def test_create_instance_group_phrase(self):
-        """Test all corner cases of _create_instance_group_phrase()"""
-        self._init_managers(models="test_model", subcommand="profile")
-
-        model_config_dict = {
-            "instance_group": [
-                {
-                    "count": 1,
-                    "kind": "KIND_GPU",
-                }
-            ]
-        }
-        self._test_instance_group_helper(model_config_dict, "1 GPU instance")
-
-        model_config_dict = {
-            "instance_group": [
-                {
-                    "count": 2,
-                    "kind": "KIND_GPU",
-                }
-            ]
-        }
-        self._test_instance_group_helper(model_config_dict, "2 GPU instances")
-
-        model_config_dict = {
-            "instance_group": [
-                {
-                    "count": 3,
-                    "kind": "KIND_GPU",
-                },
-                {
-                    "count": 2,
-                    "kind": "KIND_CPU",
-                },
-            ]
-        }
-        self._test_instance_group_helper(
-            model_config_dict, "3 GPU instances and 2 CPU instances"
-        )
-
-    def _test_instance_group_helper(self, model_config_dict, expected_output):
-        model_config = ModelConfig.create_from_dictionary(model_config_dict)
-        output = self.report_manager._create_instance_group_phrase(model_config)
-        self.assertEqual(output, expected_output)
-
-    def test_gpu_name_and_memory(self):
-        """
-        Test return value of _get_gpu_stats()
-
-        Creates a scenario where 4 GPUs are visible, but only 3 are
-        used (1 measurement uses 1 and 2, other uses 1 and 4)
-
-        The function should note that 1 and 2 are the same GPU type and combine
-        them in the name string, and it should combine the total memory of the used
-        GPUs
-        """
-        gpu_info = {
-            "gpu_uuid1": {"name": "fake_gpu_name", "total_memory": 2**30 * 2},  # 2GB
-            "gpu_uuid2": {"name": "fake_gpu_name", "total_memory": 2**30 * 2},  # 2GB
-            "gpu_uuid3": {
-                "name": "fake_gpu_name_2",
-                "total_memory": 2**30 * 4,  # 4GB
-            },
-            "gpu_uuid4": {
-                "name": "fake_gpu_name_3",
-                "total_memory": 2**30 * 8,  # 8GB
-            },
-        }
-
-        report_manager = ReportManager(
-            mode=MagicMock(),
-            config=MagicMock(),
-            gpu_info=gpu_info,
-            result_manager=MagicMock(),
-            constraint_manager=MagicMock(),
-        )
-
-        avg_gpu_metrics1 = {
-            "gpu_uuid1": {"gpu_used_memory": 6000, "gpu_utilization": 60},
-            "gpu_uuid2": {"gpu_used_memory": 6000, "gpu_utilization": 60},
-        }
-
-        avg_gpu_metrics2 = {
-            "gpu_uuid1": {"gpu_used_memory": 6000, "gpu_utilization": 60},
-            "gpu_uuid4": {"gpu_used_memory": 6000, "gpu_utilization": 60},
-        }
-
-        measurement1 = construct_run_config_measurement(
-            model_name=MagicMock(),
-            model_config_names=MagicMock(),
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values=avg_gpu_metrics1,
-            constraint_manager=MagicMock(),
-            non_gpu_metric_values=MagicMock(),
-            metric_objectives=MagicMock(),
-            model_config_weights=MagicMock(),
-        )
-
-        measurement2 = construct_run_config_measurement(
-            model_name=MagicMock(),
-            model_config_names=MagicMock(),
-            model_specific_pa_params=MagicMock(),
-            gpu_metric_values=avg_gpu_metrics2,
-            constraint_manager=MagicMock(),
-            non_gpu_metric_values=MagicMock(),
-            metric_objectives=MagicMock(),
-            model_config_weights=MagicMock(),
-        )
-
-        measurements = [measurement1, measurement2]
-        names, max_mem = report_manager._get_gpu_stats(measurements)
-        self.assertEqual(names, "2 x fake_gpu_name, 1 x fake_gpu_name_3")
-        self.assertEqual(max_mem, "12.0 GB")
-
-    @patch(
-        "model_analyzer.reports.report_manager.ReportManager._build_constraint_strings",
-        return_value={"modelA": "Max p99 latency: 100 ms"},
-    )
-    def test_constraint_string_single_model(self, *args):
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_latency_p99:
-                    max: 100
-            """
-        )
-
-        report_manager = ReportManager(
-            mode=MagicMock(),
-            config=MagicMock(),
-            gpu_info=MagicMock(),
-            result_manager=MagicMock(),
-            constraint_manager=constraint_manager,
-        )
-        expected_constraint_str = "Max p99 latency: 100 ms"
-        actual_constraint_str = report_manager._create_constraint_string(
-            report_key="modelA"
-        )
-
-        self.assertEqual(actual_constraint_str, expected_constraint_str)
-
-    @patch(
-        "model_analyzer.reports.report_manager.ReportManager._build_constraint_strings",
-        return_value={
-            "modelA": "Max p99 latency: 100 ms",
-            "modelB": "Max p99 latency: 200 ms",
-        },
-    )
-    def test_constraint_string_multi_model(self, *args):
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_latency_p99:
-                    max: 100
-              modelB:
-                constraints:
-                  perf_latency_p99:
-                    max: 200
-            """
-        )
-
-        report_manager = ReportManager(
-            mode=MagicMock(),
-            config=MagicMock(),
-            gpu_info=MagicMock(),
-            result_manager=MagicMock(),
-            constraint_manager=constraint_manager,
-        )
-        expected_constraint_str = "<strong>modelA</strong>: Max p99 latency: 100 ms"
-        expected_constraint_str += "<br>"
-
-        for i in range(len("Constraint targets: ")):
-            expected_constraint_str += "&ensp;"
-
-        expected_constraint_str += "<strong>modelB</strong>: Max p99 latency: 200 ms"
-
-        actual_constraint_str = report_manager._create_constraint_string(
-            report_key="modelA,modelB"
-        )
-
-        self.assertEqual(actual_constraint_str, expected_constraint_str)
-
-    def tearDown(self):
-        self.matplotlib_mock.stop()
-        self.io_mock.stop()
-        self.os_mock.stop()
-        self.json_mock.stop()
-        patch.stopall()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_report_utils.py b/src/c++/perf_analyzer/tests/test_report_utils.py
deleted file mode 100755
index cb22a4650..000000000
--- a/src/c++/perf_analyzer/tests/test_report_utils.py
+++ /dev/null
@@ -1,60 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.reports.report_utils import truncate_model_config_name
-
-from .common import test_result_collector as trc
-
-
-class TestReportUtils(trc.TestResultCollector):
-    """
-    Tests the report_utils functions
-    """
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_truncate_model_config_name(self):
-        """
-        Test the behavior of the truncate_longer_model_config_name function.
-        """
-
-        # Test: Shorter model config name (below 35 characters threshold)
-        model_config_name = "ensemble_model_23"
-        result = truncate_model_config_name(model_config_name)
-        self.assertEqual(model_config_name, result)
-
-        # Test: Model config name ends with 'config_#'
-        model_config_name = "long_pytorch_platform_handler_config_10"
-        result = truncate_model_config_name(model_config_name)
-        self.assertEqual(result, "long_pytorch_platform_h...config_10")
-
-        # Test: Model config name ends with 'config_default'.
-        model_config_name = "long_pytorch_platform_handler_config_default"
-        result = truncate_model_config_name(model_config_name)
-        self.assertEqual(result, "long_pytorch_platf...config_default")
-
-        # Test: Model config name includes the "config" keyword in the model name
-        model_config_name = "long_config_pytorch_platform_handler_config_128"
-        result = truncate_model_config_name(model_config_name)
-        self.assertEqual(result, "long_config_pytorch_pl...config_128")
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_result_manager.py b/src/c++/perf_analyzer/tests/test_result_manager.py
deleted file mode 100755
index a16e11b31..000000000
--- a/src/c++/perf_analyzer/tests/test_result_manager.py
+++ /dev/null
@@ -1,199 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.config.input.config_command_report import ConfigCommandReport
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.result.constraint_manager import ConstraintManager
-from model_analyzer.result.result_manager import ResultManager
-from model_analyzer.result.sorted_results import SortedResults
-from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
-
-from .common import test_result_collector as trc
-from .common.test_utils import (
-    load_multi_model_result_manager,
-    load_single_model_result_manager,
-)
-
-
-class TestResultManager(trc.TestResultCollector):
-    def test_server_data(self):
-        """
-        Test that add_server_data() and get_server_only_data()
-        are effectively mirrored set/get functions
-        """
-
-        state_manager = AnalyzerStateManager(config=MagicMock(), server=None)
-        result_manager = ResultManager(
-            config=ConfigCommandReport(),
-            state_manager=state_manager,
-            constraint_manager=ConstraintManager(config=MagicMock()),
-        )
-
-        server_data = {"a": 5, "b": 7}
-        result_manager.add_server_data(server_data)
-
-        self.assertEqual(server_data, result_manager.get_server_only_data())
-
-    def test_measurements(self):
-        """
-        Test add_run_config_measurement and get_model_configs_run_config_measurements
-
-        Confirm that run_config_measurements are stored per-model, and then per-variant, and
-        then in a list per-representation
-        Confirm that the measurements can be read out via get_model_configs_run_config_measurements()
-        """
-
-        state_manager = AnalyzerStateManager(config=MagicMock(), server=None)
-        result_manager = ResultManager(
-            config=ConfigCommandReport(),
-            state_manager=state_manager,
-            constraint_manager=ConstraintManager(config=MagicMock()),
-        )
-
-        fake_run_config1a = MagicMock()
-        fake_run_config1a.models_name.return_value = "Model1"
-        fake_run_config1a.model_variants_name.return_value = "Model1_config_default"
-        fake_run_config1a.representation.return_value = "1a"
-
-        fake_run_config1b = MagicMock()
-        fake_run_config1b.models_name.return_value = "Model1"
-        fake_run_config1b.model_variants_name.return_value = "Model1_config_default"
-        fake_run_config1b.representation.return_value = "1b"
-
-        fake_run_config1c = MagicMock()
-        fake_run_config1c.models_name.return_value = "Model1"
-        fake_run_config1c.model_variants_name.return_value = "Model1_config_0"
-        fake_run_config1c.representation.return_value = "1c"
-
-        fake_run_config2 = MagicMock()
-        fake_run_config2.models_name.return_value = "Model2"
-        fake_run_config2.model_variants_name.return_value = "Model2_config_default"
-        fake_run_config2.representation.return_value = "2"
-
-        rcm1a = MagicMock()
-        rcm1b = MagicMock()
-        rcm1c = MagicMock()
-        rcm2 = MagicMock()
-
-        result_manager._add_rcm_to_results(
-            run_config=fake_run_config1a, run_config_measurement=rcm1a
-        )
-        result_manager._add_rcm_to_results(
-            run_config=fake_run_config1b, run_config_measurement=rcm1b
-        )
-        result_manager._add_rcm_to_results(
-            run_config=fake_run_config1c, run_config_measurement=rcm1c
-        )
-        result_manager._add_rcm_to_results(
-            run_config=fake_run_config2, run_config_measurement=rcm2
-        )
-
-        config, measurements = result_manager.get_model_configs_run_config_measurements(
-            "Model1_config_default"
-        )
-        self.assertEqual(config, fake_run_config1a)
-        self.assertEqual(2, len(measurements))
-
-        config, measurements = result_manager.get_model_configs_run_config_measurements(
-            "Model1_config_0"
-        )
-        self.assertEqual(config, fake_run_config1c)
-        self.assertEqual(1, len(measurements))
-
-        config, measurements = result_manager.get_model_configs_run_config_measurements(
-            "Model2_config_default"
-        )
-        self.assertEqual(config, fake_run_config2)
-        self.assertEqual(1, len(measurements))
-
-    def test_get_single_model_names(self):
-        """Test get_model_names for a single-model run"""
-        result_manager, _ = load_single_model_result_manager()
-        self.assertEqual(result_manager.get_model_names(), ["add_sub"])
-
-    def test_get_multi_model_names(self):
-        """Test get_model_names for a multi-model run"""
-        result_manager, _ = load_multi_model_result_manager()
-        self.assertEqual(
-            result_manager.get_model_names(), ["resnet50_libtorch,vgg19_libtorch"]
-        )
-
-    def test_get_model_sorted_results_bad_name(self):
-        """
-        Test get_model_sorted_results will assert when called
-        with name that doesn't exist
-        """
-
-        result_manager, _ = load_multi_model_result_manager()
-        with self.assertRaises(TritonModelAnalyzerException):
-            result_manager.get_model_sorted_results("SHOULD_ERROR")
-
-    def test_get_single_model_sorted_results(self):
-        """
-        Test get_model_sorted_results returns a valid list
-        with only results for that model
-        """
-        result_manager, _ = load_single_model_result_manager()
-        result_manager._add_rcm_to_results(MagicMock(), MagicMock())
-        sorted_results = result_manager.get_model_sorted_results("add_sub")
-        self.assertTrue(isinstance(sorted_results, SortedResults))
-        self.assertEqual(5, len(sorted_results.results()))
-
-    def test_get_multi_model_sorted_results(self):
-        """Test get_model_sorted_results returns a valid list"""
-        result_manager, _ = load_multi_model_result_manager()
-        sorted_results = result_manager.get_model_sorted_results(
-            "resnet50_libtorch,vgg19_libtorch"
-        )
-        self.assertTrue(isinstance(sorted_results, SortedResults))
-        self.assertEqual(7, len(sorted_results.results()))
-
-    def test_get_across_model_sorted_results(self):
-        """
-        Test get_across_model_sorted_results returns a valid list
-        with all results across all models
-        """
-        result_manager, _ = load_single_model_result_manager()
-        self._add_a_fake_result(result_manager)
-
-        sorted_results = result_manager.get_across_model_sorted_results()
-        self.assertTrue(isinstance(sorted_results, SortedResults))
-        self.assertEqual(6, len(sorted_results.results()))
-
-    def _add_a_fake_result(self, result_manager):
-        fake_model = MagicMock()
-        fake_model.model_name.return_value = "FakeModel"
-        old_profile_models = result_manager._config.profile_models
-        old_profile_models.append(fake_model)
-        result_manager._config.profile_models = old_profile_models
-        result_manager._run_comparators["FakeModel"] = result_manager._run_comparators[
-            "add_sub"
-        ]
-
-        fake_run_config = MagicMock()
-        fake_run_config.models_name.return_value = "FakeModel"
-
-        result_manager.add_run_config_measurement(fake_run_config, MagicMock())
-
-    def tearDown(self):
-        patch.stopall()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_result_table.py b/src/c++/perf_analyzer/tests/test_result_table.py
deleted file mode 100755
index b7e3ee64e..000000000
--- a/src/c++/perf_analyzer/tests/test_result_table.py
+++ /dev/null
@@ -1,121 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.result.result_table import ResultTable
-
-from .common import test_result_collector as trc
-
-TEST_HEADERS = [f"Column {i}" for i in range(10)]
-TEST_COLUMN_WIDTH = 12
-TEST_WIDTHS = [TEST_COLUMN_WIDTH for i in range(10)]
-TEST_ROWS = [[f"value {i}{j}" for j in range(10)] for i in range(10)]
-TEST_TABLE_STR = (
-    "Column 0  Column 1  Column 2  Column 3  Column 4  Column 5  Column 6  Column 7  Column 8  Column 9  \n"
-    "value 00  value 01  value 02  value 03  value 04  value 05  value 06  value 07  value 08  value 09  \n"
-    "value 10  value 11  value 12  value 13  value 14  value 15  value 16  value 17  value 18  value 19  \n"
-    "value 20  value 21  value 22  value 23  value 24  value 25  value 26  value 27  value 28  value 29  \n"
-    "value 30  value 31  value 32  value 33  value 34  value 35  value 36  value 37  value 38  value 39  \n"
-    "value 40  value 41  value 42  value 43  value 44  value 45  value 46  value 47  value 48  value 49  \n"
-    "value 50  value 51  value 52  value 53  value 54  value 55  value 56  value 57  value 58  value 59  \n"
-    "value 60  value 61  value 62  value 63  value 64  value 65  value 66  value 67  value 68  value 69  \n"
-    "value 70  value 71  value 72  value 73  value 74  value 75  value 76  value 77  value 78  value 79  \n"
-    "value 80  value 81  value 82  value 83  value 84  value 85  value 86  value 87  value 88  value 89  \n"
-    "value 90  value 91  value 92  value 93  value 94  value 95  value 96  value 97  value 98  value 99  "
-)
-TEST_CSV_STR = (
-    "Column 0,Column 1,Column 2,Column 3,Column 4,Column 5,Column 6,Column 7,Column 8,Column 9\n"
-    "value 00,value 01,value 02,value 03,value 04,value 05,value 06,value 07,value 08,value 09\n"
-    "value 10,value 11,value 12,value 13,value 14,value 15,value 16,value 17,value 18,value 19\n"
-    "value 20,value 21,value 22,value 23,value 24,value 25,value 26,value 27,value 28,value 29\n"
-    "value 30,value 31,value 32,value 33,value 34,value 35,value 36,value 37,value 38,value 39\n"
-    "value 40,value 41,value 42,value 43,value 44,value 45,value 46,value 47,value 48,value 49\n"
-    "value 50,value 51,value 52,value 53,value 54,value 55,value 56,value 57,value 58,value 59\n"
-    "value 60,value 61,value 62,value 63,value 64,value 65,value 66,value 67,value 68,value 69\n"
-    "value 70,value 71,value 72,value 73,value 74,value 75,value 76,value 77,value 78,value 79\n"
-    "value 80,value 81,value 82,value 83,value 84,value 85,value 86,value 87,value 88,value 89\n"
-    "value 90,value 91,value 92,value 93,value 94,value 95,value 96,value 97,value 98,value 99"
-)
-
-
-class TestResultTableMethods(trc.TestResultCollector):
-    def tearDown(self):
-        patch.stopall()
-
-    def test_create_headers(self):
-        table = ResultTable(headers=["Column 0"])
-        self.assertEqual(table.headers(), ["Column 0"])
-        self.assertEqual(
-            table.column_widths(), [len("Column 0") + ResultTable.column_padding]
-        )
-
-    def test_add_get_methods(self):
-        table = ResultTable(headers=["Column 0", "Column 1"])
-
-        # add/get single result
-        table.insert_row_by_index(["value 0,0", "value 0,1"])
-        self.assertEqual(table.get_row_by_index(index=0), ["value 0,0", "value 0,1"])
-        table.insert_row_by_index(["value -1,0", "value -1,1"], index=0)
-        self.assertEqual(table.get_row_by_index(index=0), ["value -1,0", "value -1,1"])
-        table.insert_row_by_index(["value 1,0", "value 1,1"], index=2)
-        self.assertEqual(table.get_row_by_index(index=2), ["value 1,0", "value 1,1"])
-
-    def test_remove_methods(self):
-        table = ResultTable(headers=TEST_HEADERS)
-        # Pick a row
-        row_idx = 4
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            table.remove_row_by_index(index=row_idx)
-
-        # Fill table
-        for row in TEST_ROWS:
-            table.insert_row_by_index(row)
-
-        self.assertEqual(
-            table.get_row_by_index(index=row_idx), [f"value 4{j}" for j in range(10)]
-        )
-
-        # remove rows and check that the next one is in its place
-        table.remove_row_by_index(index=row_idx)
-        self.assertEqual(
-            table.get_row_by_index(index=row_idx), [f"value 5{j}" for j in range(10)]
-        )
-
-        # Remove all rows and then one
-        with self.assertRaises(TritonModelAnalyzerException):
-            for i in range(10):
-                table.remove_row_by_index(index=0)
-
-    def test_to_formatted_string(self):
-        table = ResultTable(headers=TEST_HEADERS)
-        for row in TEST_ROWS:
-            table.insert_row_by_index(row)
-        self.assertEqual(table.to_formatted_string(), TEST_TABLE_STR)
-
-        table = ResultTable(headers=TEST_HEADERS)
-        for row in TEST_ROWS:
-            table.insert_row_by_index(row)
-        self.assertEqual(
-            table.to_formatted_string(separator=",", ignore_widths=True), TEST_CSV_STR
-        )
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_result_table_manager.py b/src/c++/perf_analyzer/tests/test_result_table_manager.py
deleted file mode 100755
index 278ceb5ea..000000000
--- a/src/c++/perf_analyzer/tests/test_result_table_manager.py
+++ /dev/null
@@ -1,344 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from filecmp import cmp
-from shutil import rmtree
-from unittest.mock import patch
-
-from google.protobuf import text_format
-from tritonclient.grpc import model_config_pb2
-
-from model_analyzer.result.constraint_manager import ConstraintManager
-from model_analyzer.result.result_manager import ResultManager
-from model_analyzer.result.result_table_manager import ResultTableManager
-from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
-
-from .common import test_result_collector as trc
-from .common.test_utils import (
-    ROOT_DIR,
-    evaluate_mock_config,
-    load_bls_result_manager,
-    load_ensemble_result_manager,
-    load_multi_model_result_manager,
-    load_request_rate_result_manager,
-    load_single_model_result_manager,
-)
-
-
-class TestResultTableManager(trc.TestResultCollector):
-    def test_single_model_csv_against_golden(self):
-        """
-        Match the csvs against the golden versions in
-        tests/common/single-model-ckpt
-        """
-        table_manager = self._create_single_model_result_table_manager()
-
-        table_manager.create_tables()
-        table_manager.tabulate_results()
-        table_manager.export_results()
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/single-model-ckpt/results/metrics-model-gpu.csv",
-                f"{ROOT_DIR}/single-model-ckpt/golden-metrics-model-gpu.csv",
-            )
-        )
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/single-model-ckpt/results/metrics-model-inference.csv",
-                f"{ROOT_DIR}/single-model-ckpt/golden-metrics-model-inference.csv",
-            )
-        )
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/single-model-ckpt/results/metrics-server-only.csv",
-                f"{ROOT_DIR}/single-model-ckpt/golden-metrics-server-only.csv",
-            )
-        )
-
-        rmtree(f"{ROOT_DIR}/single-model-ckpt/results/")
-
-    def test_multi_model_csv_against_golden(self):
-        """
-        Match the csvs against the golden versions in
-        tests/common/multi-model-ckpt
-        """
-        table_manager = self._create_multi_model_result_table_manager()
-
-        table_manager.create_tables()
-        table_manager.tabulate_results()
-        table_manager.export_results()
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/multi-model-ckpt/results/metrics-model-gpu.csv",
-                f"{ROOT_DIR}/multi-model-ckpt/golden-metrics-model-gpu.csv",
-            )
-        )
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/multi-model-ckpt/results/metrics-model-inference.csv",
-                f"{ROOT_DIR}/multi-model-ckpt/golden-metrics-model-inference.csv",
-            )
-        )
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/multi-model-ckpt/results/metrics-server-only.csv",
-                f"{ROOT_DIR}/multi-model-ckpt/golden-metrics-server-only.csv",
-            )
-        )
-
-        rmtree(f"{ROOT_DIR}/multi-model-ckpt/results/")
-
-    def test_ensemble_csv_against_golden(self):
-        """
-        Match the csvs against the golden versions in
-        tests/common/ensemble-ckpt
-        """
-        table_manager = self._create_ensemble_result_table_manager()
-
-        table_manager.create_tables()
-        table_manager.tabulate_results()
-        table_manager.export_results()
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/ensemble-ckpt/results/metrics-model-gpu.csv",
-                f"{ROOT_DIR}/ensemble-ckpt/golden-metrics-model-gpu.csv",
-            )
-        )
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/ensemble-ckpt/results/metrics-model-inference.csv",
-                f"{ROOT_DIR}/ensemble-ckpt/golden-metrics-model-inference.csv",
-            )
-        )
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/ensemble-ckpt/results/metrics-server-only.csv",
-                f"{ROOT_DIR}/ensemble-ckpt/golden-metrics-server-only.csv",
-            )
-        )
-
-        rmtree(f"{ROOT_DIR}/ensemble-ckpt/results/")
-
-    def test_request_rate_csv_against_golden(self):
-        """
-        Match the csvs against the golden versions in
-        tests/common/request-rate-ckpt
-        """
-        table_manager = self._create_request_rate_result_table_manager()
-
-        table_manager.create_tables()
-        table_manager.tabulate_results()
-        table_manager.export_results()
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/request-rate-ckpt/results/metrics-model-gpu.csv",
-                f"{ROOT_DIR}/request-rate-ckpt/golden-metrics-model-gpu.csv",
-            )
-        )
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/request-rate-ckpt/results/metrics-model-inference.csv",
-                f"{ROOT_DIR}/request-rate-ckpt/golden-metrics-model-inference.csv",
-            )
-        )
-
-        self.assertTrue(
-            cmp(
-                f"{ROOT_DIR}/request-rate-ckpt/results/metrics-server-only.csv",
-                f"{ROOT_DIR}/request-rate-ckpt/golden-metrics-server-only.csv",
-            )
-        )
-
-        rmtree(f"{ROOT_DIR}/request-rate-ckpt/results/")
-
-    def test_create_inference_table_with_backend_parameters(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "-f",
-            "config.yml",
-            "--model-repository",
-            ".",
-        ]
-        yaml_str = """
-            profile_models: profile_models
-            inference_output_fields: model_name,batch_size,backend_parameter/parameter_1,backend_parameter/parameter_2
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-        state_manager = AnalyzerStateManager(config=config, server=None)
-        result_manager = ResultManager(
-            config=config,
-            state_manager=state_manager,
-            constraint_manager=ConstraintManager(config),
-        )
-        result_table_manager = ResultTableManager(
-            config=config, result_manager=result_manager
-        )
-
-        result_table_manager._create_inference_table()
-        self.assertEqual(
-            result_table_manager._inference_output_fields,
-            [
-                "model_name",
-                "batch_size",
-                "backend_parameter/parameter_1",
-                "backend_parameter/parameter_2",
-            ],
-        )
-
-    def test_get_common_row_items_with_backend_parameters(self):
-        """
-        This tests that a metrics model inference table row can be created with
-        backend parameters included. Each backend parameter gets its own column.
-        The column name is the backend parameter key (prepended with a prefix
-        to avoid potentially overlapping with an existing column). The column
-        value is the backend parameter value.
-
-        Here is an example table:
-
-        Models (Inference):
-        Model     Model Config Path   backend_parameter/add_sub_key_1   backend_parameter/add_sub_key_2
-        add_sub   add_sub_config_2    add_sub_value_1                   add_sub_value_2
-        add_sub   add_sub_config_0    add_sub_value_1                   add_sub_value_2
-        add_sub   add_sub_config_1    add_sub_value_1                   add_sub_value_2
-
-        Each row of the metrics model inference table corresponds to one model
-        config variant.
-
-        It is possible for a user to run the profile command with multiple
-        models config variants from different models with potentially different
-        backend parameters. This test includes backend parameters from two
-        separate models, showing that for one particular row (for a 'model A'
-        config variant), it only populates the backend parameter cells for
-        'model A', and the backend parameter cells for 'model B' are empty
-        (None).
-
-        Here is an example table with backend parameters from different models:
-
-        Models (Inference):
-        Model       Model Config Path   backend_parameter/add_sub_key_1   backend_parameter/add_sub_key_2   backend_parameter/add_sub_2_key_1   backend_parameter/add_sub_2_key_2
-        add_sub     add_sub_config_2    add_sub_value_1                   add_sub_value_2                   None                                None
-        add_sub     add_sub_config_0    add_sub_value_1                   add_sub_value_2                   None                                None
-        add_sub     add_sub_config_1    add_sub_value_1                   add_sub_value_2                   None                                None
-        add_sub_2   add_sub_2_config_2  None                              None                              add_sub_2_value_1                   add_sub_2_value_2
-        add_sub_2   add_sub_2_config_1  None                              None                              add_sub_2_value_1                   add_sub_2_value_2
-        add_sub_2   add_sub_2_config_0  None                              None                              add_sub_2_value_1                   add_sub_2_value_2
-        """
-
-        args = [
-            "model-analyzer",
-            "profile",
-            "-f",
-            "config.yml",
-            "--model-repository",
-            ".",
-        ]
-        yaml_str = """
-            profile_models: profile_models
-            inference_output_fields: model_name,batch_size,backend_parameter/model_1_key_1,backend_parameter/model_1_key_2,backend_parameter/model_2_key_1
-        """
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-        state_manager = AnalyzerStateManager(config=config, server=None)
-        result_manager = ResultManager(
-            config=config,
-            state_manager=state_manager,
-            constraint_manager=ConstraintManager(config),
-        )
-        table_manager = ResultTableManager(config=config, result_manager=result_manager)
-        model_config_str = """
-            parameters: {
-            key: "model_1_key_1"
-                value: {
-                string_value:"model_1_value_1"
-                }
-            }
-            parameters: {
-            key:"model_1_key_2"
-                value: {
-                string_value:"model_1_value_2"
-                }
-            }
-            """
-        backend_parameters = text_format.Parse(
-            model_config_str, model_config_pb2.ModelConfig()
-        ).parameters
-        row = table_manager._get_common_row_items(
-            fields=[
-                "model_name",
-                "batch_size",
-                "backend_parameter/model_1_key_1",
-                "backend_parameter/model_1_key_2",
-                "backend_parameter/model_2_key_1",
-            ],
-            batch_sizes="batch_size",
-            concurrencies=None,
-            request_rates=None,
-            satisfies=None,
-            model_name="model_name",
-            model_config_path=None,
-            dynamic_batchings=None,
-            instance_groups=None,
-            max_batch_sizes=None,
-            backend_parameters=backend_parameters,
-        )
-        self.assertEqual(
-            row,
-            ["model_name", "batch_size", "model_1_value_1", "model_1_value_2", None],
-        )
-
-    def _create_single_model_result_table_manager(self):
-        result_manager, config = load_single_model_result_manager()
-
-        return ResultTableManager(config, result_manager)
-
-    def _create_multi_model_result_table_manager(self):
-        result_manager, config = load_multi_model_result_manager()
-
-        return ResultTableManager(config, result_manager)
-
-    def _create_ensemble_result_table_manager(self):
-        result_manager, config = load_ensemble_result_manager()
-
-        return ResultTableManager(config, result_manager)
-
-    def _create_bls_result_table_manager(self):
-        result_manager, config = load_bls_result_manager()
-
-        return ResultTableManager(config, result_manager)
-
-    def _create_request_rate_result_table_manager(self):
-        result_manager, config = load_request_rate_result_manager()
-
-        return ResultTableManager(config, result_manager)
-
-    def tearDown(self):
-        patch.stopall()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_run_config_generator.py b/src/c++/perf_analyzer/tests/test_run_config_generator.py
deleted file mode 100755
index b58850b9f..000000000
--- a/src/c++/perf_analyzer/tests/test_run_config_generator.py
+++ /dev/null
@@ -1,729 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.config.generate.brute_run_config_generator import (
-    BruteRunConfigGenerator,
-)
-from model_analyzer.config.generate.generator_utils import GeneratorUtils as utils
-from model_analyzer.config.generate.model_profile_spec import ModelProfileSpec
-from model_analyzer.config.generate.model_run_config_generator import (
-    ModelRunConfigGenerator,
-)
-from model_analyzer.config.generate.model_variant_name_manager import (
-    ModelVariantNameManager,
-)
-from model_analyzer.config.input.config_defaults import (
-    DEFAULT_RUN_CONFIG_MAX_CONCURRENCY,
-    DEFAULT_RUN_CONFIG_MAX_INSTANCE_COUNT,
-    DEFAULT_RUN_CONFIG_MAX_MODEL_BATCH_SIZE,
-)
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-
-from .common import test_result_collector as trc
-from .common.test_utils import construct_run_config_measurement, evaluate_mock_config
-from .mocks.mock_model_config import MockModelConfig
-from .mocks.mock_os import MockOSMethods
-
-
-class TestRunConfigGenerator(trc.TestResultCollector):
-    def __init__(self, methodname):
-        super().__init__(methodname)
-        self._fake_throughput = 1
-
-    def test_default_config_single_model(self):
-        """
-        Test Default Single Model:
-
-        num_PAC = log2(DEFAULT_RUN_CONFIG_MAX_CONCURRENCY) + 1
-        num_MC = (  DEFAULT_RUN_CONFIG_MAX_INSTANCE_COUNT
-                  x log2(DEFAULT_RUN_CONFIG_MAX_MODEL_BATCH_SIZE)
-                 ) + 1
-        total = (num_PAC * num_MC) will be generated by the auto-search
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            profile_models:
-                - my-model
-            """)
-        # yapf: enable
-
-        expected_pa_configs = len(
-            utils.generate_doubled_list(1, DEFAULT_RUN_CONFIG_MAX_CONCURRENCY)
-        )
-
-        expected_model_configs = (
-            DEFAULT_RUN_CONFIG_MAX_INSTANCE_COUNT
-            * len(
-                utils.generate_doubled_list(1, DEFAULT_RUN_CONFIG_MAX_MODEL_BATCH_SIZE)
-            )
-            + 1
-        )
-        expected_num_of_configs = expected_pa_configs * expected_model_configs
-
-        self._run_and_test_run_config_generator(
-            yaml_str, expected_config_count=expected_num_of_configs
-        )
-
-    def test_two_models(self):
-        """
-        Test Two Models that have the same automatic configuration:
-
-        num_PAC = 2
-        num_MC = 2 * 2 = 4
-        model_total = (2 * 4) = 8 will be generated by the auto-search for each model
-
-        default_step = (2 * 2) = 4 will be generated by combining the default configs (2 for each PAC)
-
-        total = default_step + model_total * model_total = 68
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 2
-            run_config_search_max_instance_count: 2
-            run_config_search_max_concurrency: 2
-            profile_models:
-                - my-model
-                - my-modelB
-
-            """)
-        # yapf: enable
-
-        expected_num_of_configs = 68
-
-        # Expect 78 calls to ModelRunConfigGenerator.set_last_results
-        # All 68 experiments will be passed to the leaf generator
-        # All 2 times that the leaf generator is done with default config will also pass results to root generator
-        # All 8 times that the leaf generator is done with non-default config will also pass results to root generator
-
-        expected_num_calls_to_set_last_results = 78
-
-        with patch.object(
-            ModelRunConfigGenerator,
-            "set_last_results",
-            side_effect=ModelRunConfigGenerator.set_last_results,
-            autospec=True,
-        ) as mock_method:
-            self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-            self.assertEqual(
-                mock_method.call_count, expected_num_calls_to_set_last_results
-            )
-
-    def test_two_uneven_models(self):
-        """
-        Test Two Uneven Models:
-
-        Model A is auto search:
-            num_PAC = 3
-            num_MC = 2 * 2 = 4
-            modelA_total = (3 * 4) = 12
-
-        Model B is manual search:
-            num_PAC = 2
-            num_MC = 2 * 3 = 6
-            modelB_total = 2 * 6 = 12
-
-        default_step = num_PAC_A * num_PAC_B = 6
-        total = default_step + modelA_total * modelB_total = 150
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 2
-            run_config_search_max_instance_count: 2
-            run_config_search_max_concurrency: 2
-            profile_models:
-                my-model:
-                    parameters:
-                        concurrency: [1,2,3]
-                my-modelB:
-                    model_config_parameters:
-                        max_batch_size: [1,4,16]
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: [1,2]
-            """)
-        # yapf: enable
-
-        expected_num_of_configs = 150
-
-        self._run_and_test_run_config_generator(
-            yaml_str, expected_config_count=expected_num_of_configs
-        )
-
-    def test_three_uneven_models(self):
-        """
-        Test Three Uneven Models:
-
-        Model A is auto search:
-            num_PAC = 2
-            num_MC = 2 * 2 = 4
-            modelA_total = 2 * 4 = 8
-
-        Model B is auto search with fixed concurrency:
-            num_PAC = 3
-            num_MC = 2 * 2 = 4
-            modelB_total = 3 *4 = 12
-
-        Model C is manual search:
-            num_PAC = 2
-            num_MC = 3 * 2 = 6
-            modelC_total = 2 * 6 = 12
-
-        default_step = num_PAC_A * num_PAC_B * num_PAC_C = 2*3*2 = 12
-
-        total = default_step + modelA_total * modelB_total * modelC_total = 1164
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 2
-            run_config_search_max_instance_count: 2
-            run_config_search_max_concurrency: 2
-            profile_models:
-                - my-model
-                -
-                  my-modelB:
-                    parameters:
-                        concurrency: [1,5,9]
-                -
-                  my-model3:
-                    parameters:
-                        concurrency: [10, 20]
-                    model_config_parameters:
-                        max_batch_size: [1,64]
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: [1,2,3]
-            """)
-        # yapf: enable
-
-        expected_num_of_configs = 1164
-        # All 2 times that the leaf generator is done with default config will also pass results to root generator
-        # All 8 times that the leaf generator is done with non-default config will also pass results to root generator
-
-        # Expect 1268 calls to ModelRunConfigGenerator.set_last_results
-        # All 1164 experiments will be passed to the leaf generator
-        # All 96 times that the leaf generator is done with non-default config will pass results to the middle generator
-        # All 6 times that the leaf generator is done with default config will pass results to the middle generator
-        # All 8 times that the middle generator is done with non-default config will pass results to the root generator
-        # All 2 times that the middle generator is done with default config will pass results to the root generator
-        expected_num_calls_to_set_last_results = 1276
-
-        with patch.object(
-            ModelRunConfigGenerator,
-            "set_last_results",
-            side_effect=ModelRunConfigGenerator.set_last_results,
-            autospec=True,
-        ) as mock_method:
-            self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-            self.assertEqual(
-                mock_method.call_count, expected_num_calls_to_set_last_results
-            )
-
-    def test_early_backoff_leaf_model(self):
-        """
-        Test the case where there are two models, and the 'leaf' model (the last generator called
-        in the recursive generator stack) has early backoff
-
-        Both models are auto search:
-            num_PAC = 2
-            num_MC = 4 * 2 = 8
-            model_total = 2 * 8 = 16
-
-        default_step = num_PAC_A * num_PAC_B = 4
-        total = default_step + modelA_total * modelA_total = 260
-
-        However, the test will set up the throughput values such that ModelB will see a lack
-        of throughput the first time that it is walking max_batch_size.
-        Normally it would walk values 1,2,4,8, but for this ONE case it will only walk 1,2.
-        This will reduce the total count by 4, as there are two concurrencies that would be
-        tried for ModelB for each of the two removed max_batch_sizes
-
-        Thus, actual expected_count = 256
-
-
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 8
-            run_config_search_max_instance_count: 2
-            run_config_search_max_concurrency: 2
-            profile_models:
-                - my-model
-                - my-modelB
-            """)
-        # yapf: enable
-
-        expected_num_of_configs = 256
-
-        perf_throughput_values = [2**i for i in range(expected_num_of_configs)]
-        perf_throughput_values[6] = perf_throughput_values[5]
-        perf_throughput_values[7] = perf_throughput_values[5]
-
-        with patch.object(
-            TestRunConfigGenerator, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.side_effect = perf_throughput_values
-            self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-    def test_early_backoff_root_model(self):
-        """
-        Test the case where there are two models, and the 'root' model (the first one
-        called in the recursive generator stack) has early backoff
-
-        Both models are auto search:
-            num_PAC = 2
-            num_MC = 4 * 2 = 8
-            model_total = 2 * 8 = 16
-
-        default_step = num_PAC_A * num_PAC_B = 4
-        total = default_step + modelA_total * modelA_total = 260
-
-        However, the test will set up the throughput values such that ModelA will see a lack
-        of throughput the first time that it is walking max_batch_size.
-        Normally it would walk values 1,2,4,8, but for this case it will only walk 1,2.
-        This will reduce the total count by 64: (2 max_batch_sizes * 2 concurrencies * 16 modelB cases)
-
-        Thus, actual expected_count = 196
-
-
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 8
-            run_config_search_max_instance_count: 2
-            run_config_search_max_concurrency: 2
-            profile_models:
-                - my-model
-                - my-modelB
-            """)
-        # yapf: enable
-
-        expected_num_of_configs = 196
-
-        perf_throughput_values = [2**i for i in range(expected_num_of_configs)]
-        # First 4 is modelA=default, modelB=default
-        # Next 32 is modelA max_batch_size=1, modelB=all 16 cases
-        # Next 32 is modelA max_batch_size=2, modelB=all 16 cases. We want to change these to show no increase
-        for i in range(36, 68):
-            perf_throughput_values[i] = perf_throughput_values[i - 36]
-
-        with patch.object(
-            TestRunConfigGenerator, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.side_effect = perf_throughput_values
-            self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-    def test_measurement_list(self):
-        """
-        Test that the root model (the first one called in the recursive generator stack) gets a list
-        of all measurements since the last time it took a step, and makes a decision based on the
-        maximum throughput observed, not just the last measurement
-
-        Both models are auto search:
-            num_PAC = 2
-            num_MC = 4 * 2 = 8
-            model_total = 2 * 8 = 16
-
-        default_total = num_PAC_A * num_PAC_B = 2*2 = 4
-        total = default_total + modelA_total * modelB_total = 324
-
-        The test will set up the throughput values such that there is an increase to the maximum throughput
-        between each step of the root model, but that those results aren't strictly increasing every time.
-
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 8
-            run_config_search_max_instance_count: 2
-            run_config_search_max_concurrency: 2
-            profile_models:
-                - my-model
-                - my-modelB
-            """)
-        # yapf: enable
-
-        expected_num_of_configs = 260
-
-        # The first 4 results (0-3) are modelA=default, modelB=default
-        # The next 32 results (4-35) are modelA= max_batch_size=1 concurrency 1 and 2, modelB all 16 combinations
-        # The next 32 results (36-67) are modelA= max_batch_size=2 concurrency 1 and 2, modelB all 16 combinations
-        #
-        # Result 67 is the one to change to a smaller result for this test to work. If modelA was incorrectly
-        # only looking at 67 instead of all results 36-67, then it would incorrectly cut off the max_batch_size
-        # search and would not end up returning a full 260 configs
-        perf_throughput_values = list(range(260))
-        perf_throughput_values[67] = 1
-
-        with patch.object(
-            TestRunConfigGenerator, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.side_effect = perf_throughput_values
-            self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-    def test_matching_triton_server_env(self):
-        """
-        Test that we don't assert if triton server environments match:
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 2
-            run_config_search_max_instance_count: 2
-            run_config_search_max_concurrency: 2
-            profile_models:
-                -
-                  my-model:
-                    triton_server_environment:
-                        'LD_PRELOAD': fake_preload_1,
-                        'LD_LIBRARY_PATH': fake_library_path_1
-                -
-                  my-modelB:
-                    triton_server_environment:
-                        'LD_PRELOAD': fake_preload_1,
-                        'LD_LIBRARY_PATH': fake_library_path_1
-            """)
-        # yapf: enable
-
-        expected_num_of_configs = 68
-        self._run_and_test_run_config_generator(
-            yaml_str, expected_config_count=expected_num_of_configs
-        )
-
-    def test_mismatching_triton_server_env(self):
-        """
-        Test that we assert if triton server environments don't match:
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 2
-            run_config_search_max_instance_count: 2
-            run_config_search_max_concurrency: 2
-            profile_models:
-                -
-                  my-model:
-                    triton_server_environment:
-                        'LD_PRELOAD': fake_preload_1,
-                        'LD_LIBRARY_PATH': fake_library_path_1
-                -
-                  my-modelB:
-                    triton_server_environment:
-                        'LD_PRELOAD': fake_preload_2,
-                        'LD_LIBRARY_PATH': fake_library_path_2
-            """)
-        # yapf: enable
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            expected_num_of_configs = 100
-            self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-    def test_none_result_in_max_batch_size(self):
-        """
-        Test that if we get None results for an entire set of MaxBatch experiments that
-        we properly stop stepping max batch size and do not assert
-
-        When the test steps to 2 instances with max_batch_size of 2, it will return None as the results.
-        This should stop the entire max_batch_size walk for 2 instances, and should step to 3 instances
-        """
-
-        expected_num_of_configs = 17
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 4
-            run_config_search_max_instance_count: 3
-            run_config_search_max_concurrency: 2
-            profile_models:
-                - my-model
-            """)
-
-        perf_throughput_values = [
-            1, 2,  # Default, Concurrency 1 and 2
-            1, 2,  # Instances=1, MaxBatch=1, Concurrency 1 and 2
-            3, 4,  # Instances=1, MaxBatch=2, Concurrency 1 and 2
-            5, 6,  # Instances=1, MaxBatch=4, Concurrency 1 and 2
-            7, 8,  # Instances=2, MaxBatch=1, Concurrency 1 and 2
-            None,  # Instances=2, MaxBatch=2, Concurrency 1.
-            9,10,  # Instances=3, MaxBatch=1, Concurrency 1 and 2
-            11,12, # Instances=3, MaxBatch=2, Concurrency 1 and 2
-            13,14  # Instances=3, MaxBatch=4, Concurrency 1 and 2
-        ]
-        # yapf: enable
-
-        with patch.object(
-            TestRunConfigGenerator, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.side_effect = perf_throughput_values
-            self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-    def test_none_result_before_threshold(self):
-        """
-        Test that a 'none' result (from PA erroring) before the minimum number
-        of throughput runs (4) will properly early exit and not assert
-        """
-
-        # Full sweep is 3 model configs * 12 concurrencies = 36.
-        # However, we will early exit and skip 9 of the concurrencies
-        # Thus, expect 27
-        #
-        expected_num_of_configs = 27
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 2
-            run_config_search_max_instance_count: 1
-            run_config_search_max_concurrency: 2048
-            profile_models:
-                - my-model
-            """)
-
-        perf_throughput_values = [
-            1,2,4,8,16,32,64,128,256,512,1024,2048, # Default config
-            1,2,None,                               # Batch size 1
-            1,2,4,8,16,32,64,128,256,512,1024,2048  # Batch size 2
-        ]
-        # yapf: enable
-
-        with patch.object(
-            TestRunConfigGenerator, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.side_effect = perf_throughput_values
-            self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-    def test_none_result_after_threshold(self):
-        """
-        Test that a 'none' result (from PA erroring) after the minimum number
-        of throughput runs (4) will properly early exit and not assert
-        """
-
-        # Full sweep is 3 model configs * 12 concurrencies = 36.
-        # However, we will early exit and skip 1 of the concurrencies
-        # Thus, expect 35
-        #
-        expected_num_of_configs = 35
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 2
-            run_config_search_max_instance_count: 1
-            run_config_search_max_concurrency: 2048
-            profile_models:
-                - my-model
-            """)
-
-        perf_throughput_values = [
-            1,2,4,8,16,32,64,128,256,512,1024,2048, # Default config
-            1,2,4,8,16,32,64,128,256,512,None,      # Batch size 1
-            1,2,4,8,16,32,64,128,256,512,1024,2048  # Batch size 2
-        ]
-        # yapf: enable
-
-        with patch.object(
-            TestRunConfigGenerator, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.side_effect = perf_throughput_values
-            self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-    def test_variant_naming(self):
-        """
-        Confirm that model variant names are consistent
-
-        The test is set up such that on the first pass, we will skip
-        model B's case of 1 instance, batch size=4. On the second pass
-        (when the root model max_batch_size of 2), no cases will be skipped.
-        As a result, the naming will be this for model B:
-            modelB_config_0: 1 instance, max_batch_size=1
-            modelB_config_1: 1 instance, max_batch_size=2
-            modelB_config_5: 1 instance, max_batch_size=4
-            modelB_config_2: 2 instance, max_batch_size=1
-            modelB_config_3: 2 instance, max_batch_size=2
-            modelB_config_4: 2 instance, max_batch_size=4
-        """
-
-        # yapf: disable
-        yaml_str = ("""
-            run_config_search_max_model_batch_size: 4
-            run_config_search_max_instance_count: 2
-            run_config_search_max_concurrency: 1
-
-            profile_models:
-                -
-                  modelA:
-                    model_config_parameters:
-                        max_batch_size: [1,2]
-                        instance_group:
-                        -
-                            kind: KIND_GPU
-                            count: [1]
-                - modelB
-            """)
-
-        expected_modelB_name_order = [
-            "modelB_config_default",
-            "modelB_config_0",
-            "modelB_config_1",
-            "modelB_config_2",
-            "modelB_config_3",
-            "modelB_config_4",
-            "modelB_config_0",
-            "modelB_config_1",
-            "modelB_config_5", # This was skipped, and needs to be created on second pass
-            "modelB_config_2",
-            "modelB_config_3",
-            "modelB_config_4",
-        ]
-
-        perf_throughput_values = [
-            1,      # Default config
-            1,None, # A: 1 Instance, BS=1  B: 1 Instance, BS=1,2
-            2,3,4,  # A: 1 Instance, BS=1  B: 2 Instance, BS=1,2,4
-            5,6,7,  # A: 1 Instance, BS=2  B: 1 Instance, BS=1,2,4
-            8,9,10  # A: 1 Instance, BS=2  B: 2 Instance, BS=1,2,4
-        ]
-        # yapf: enable
-
-        expected_num_of_configs = 12
-
-        with patch.object(
-            TestRunConfigGenerator, "_get_next_perf_throughput_value"
-        ) as mock_method:
-            mock_method.side_effect = perf_throughput_values
-            run_configs = self._run_and_test_run_config_generator(
-                yaml_str, expected_config_count=expected_num_of_configs
-            )
-
-        for i, rc in enumerate(run_configs):
-            self.assertEqual(
-                expected_modelB_name_order[i],
-                rc.model_run_configs()[1].model_variant_name(),
-            )
-
-    def _run_and_test_run_config_generator(self, yaml_str, expected_config_count):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-        ]
-
-        protobuf = """
-            max_batch_size: 8
-            instance_group [
-            {
-                kind: KIND_CPU
-                count: 1
-            }
-            ]
-            """
-
-        self.mock_model_config = MockModelConfig(protobuf)
-        self.mock_model_config.start()
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        profile_models = []
-        for model in config.profile_models:
-            profile_models.append(
-                ModelProfileSpec(model, config, MagicMock(), MagicMock())
-            )
-
-        rcg = BruteRunConfigGenerator(
-            config, MagicMock(), profile_models, MagicMock(), ModelVariantNameManager()
-        )
-
-        run_configs = []
-        for run_config in rcg.get_configs():
-            run_configs.append(run_config)
-            rcg.set_last_results(self._get_next_fake_results())
-
-        self.assertEqual(expected_config_count, len(set(run_configs)))
-
-        # Checks that each ModelRunConfig contains the expected number of model_configs
-        for run_config in run_configs:
-            self.assertEqual(
-                len(run_config.model_run_configs()), len(config.profile_models)
-            )
-
-        self.mock_model_config.stop()
-
-        return run_configs
-
-    def setUp(self):
-        # Mock path validation
-        self.mock_os = MockOSMethods(
-            mock_paths=["model_analyzer.config.input.config_utils"]
-        )
-        self.mock_os.start()
-
-    def tearDown(self):
-        self.mock_os.stop()
-        patch.stopall()
-
-    def _get_next_fake_results(self):
-        throughput_value = self._get_next_perf_throughput_value()
-
-        measurement = None
-
-        if throughput_value is not None:
-            measurement = construct_run_config_measurement(
-                model_name=MagicMock(),
-                model_config_names=["test_config_name"],
-                model_specific_pa_params=MagicMock(),
-                gpu_metric_values=MagicMock(),
-                non_gpu_metric_values=[{"perf_throughput": throughput_value}],
-            )
-
-        return [measurement]
-
-    def _get_next_perf_throughput_value(self):
-        self._fake_throughput *= 2
-        return self._fake_throughput
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_run_config_measurement.py b/src/c++/perf_analyzer/tests/test_run_config_measurement.py
deleted file mode 100755
index 579af9e2a..000000000
--- a/src/c++/perf_analyzer/tests/test_run_config_measurement.py
+++ /dev/null
@@ -1,739 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import json
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.result.run_config_measurement import RunConfigMeasurement
-from tests.common.test_utils import (
-    construct_constraint_manager,
-    construct_run_config_measurement,
-    convert_avg_gpu_metrics_to_data,
-    convert_gpu_metrics_to_data,
-    convert_non_gpu_metrics_to_data,
-    default_encode,
-)
-
-from .common import test_result_collector as trc
-
-
-class TestRunConfigMeasurement(trc.TestResultCollector):
-    def setUp(self):
-        self._construct_rcm0()
-        self._construct_rcm1()
-        self._construct_rcm2()
-        self._construct_rcm3()
-        self._construct_rcm4()
-        self._construct_rcm5()
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_model_variants_name(self):
-        """
-        Test that the model_variants_name was initialized correctly
-        """
-        self.assertEqual(self.rcm0.model_variants_name(), self.model_variants_name)
-
-    def test_gpu_data(self):
-        """
-        Test that the gpu data is correct
-        """
-        self.assertEqual(
-            self.rcm0.gpu_data(), convert_gpu_metrics_to_data(self.gpu_metric_values)
-        )
-
-    def test_non_gpu_data(self):
-        """
-        Test that the non-gpu data is correct
-        """
-        self.assertEqual(
-            self.rcm0.non_gpu_data(),
-            [
-                convert_non_gpu_metrics_to_data(ngvm)
-                for ngvm in self.rcm0_non_gpu_metric_values
-            ],
-        )
-
-    def test_data(self):
-        """
-        Test that the gpu + non-gpu data is correct
-        """
-        avg_gpu_data = convert_avg_gpu_metrics_to_data(self.avg_gpu_metric_values)
-
-        data = [
-            list(avg_gpu_data.values()) + convert_non_gpu_metrics_to_data(ngvm)
-            for ngvm in self.rcm0_non_gpu_metric_values
-        ]
-
-        self.assertEqual(list(self.rcm0.data().values()), data)
-
-    def test_gpus_used(self):
-        """
-        Test that the list of gpus used is correct
-        """
-        self.assertEqual(self.rcm0.gpus_used(), ["0", "1"])
-
-    def test_get_gpu_metric(self):
-        """
-        Test that the gpu metric data is correct
-        """
-        avg_gpu_data = convert_avg_gpu_metrics_to_data(self.avg_gpu_metric_values)
-
-        self.assertEqual(
-            self.rcm0.get_gpu_metric("gpu_used_memory"), avg_gpu_data["gpu_used_memory"]
-        )
-
-        self.assertEqual(
-            self.rcm0.get_gpu_metric("gpu_utilization"), avg_gpu_data["gpu_utilization"]
-        )
-
-    def test_gpu_gpu_metric_not_found(self):
-        """
-        Test that an incorrect metric search returns None
-        """
-        self.assertEqual(self.rcm0.get_gpu_metric("XXXXX"), None)
-
-    def test_get_non_gpu_metric(self):
-        """
-        Test that the non-gpu metric data is correct
-        """
-        non_gpu_data = [
-            convert_non_gpu_metrics_to_data(non_gpu_metric_value)
-            for non_gpu_metric_value in self.rcm0_non_gpu_metric_values
-        ]
-
-        self.assertEqual(
-            self.rcm0.get_non_gpu_metric("perf_throughput"),
-            [non_gpu_data[0][0], non_gpu_data[1][0]],
-        )
-        self.assertEqual(
-            self.rcm0.get_non_gpu_metric("perf_latency_p99"),
-            [non_gpu_data[0][1], non_gpu_data[1][1]],
-        )
-        self.assertEqual(
-            self.rcm0.get_non_gpu_metric("cpu_used_ram"),
-            [non_gpu_data[0][2], non_gpu_data[1][2]],
-        )
-
-    def test_get_weighted_non_gpu_metric(self):
-        """
-        Test that the weighted non-gpu metric data is correct
-        """
-        non_gpu_data = [
-            convert_non_gpu_metrics_to_data(weighted_non_gpu_metric_value)
-            for weighted_non_gpu_metric_value in self.rcm0_weighted_non_gpu_metric_values
-        ]
-
-        self.assertEqual(
-            self.rcm0.get_weighted_non_gpu_metric("perf_throughput"),
-            [non_gpu_data[0][0], non_gpu_data[1][0]],
-        )
-        self.assertEqual(
-            self.rcm0.get_weighted_non_gpu_metric("perf_latency_p99"),
-            [non_gpu_data[0][1], non_gpu_data[1][1]],
-        )
-        self.assertEqual(
-            self.rcm0.get_weighted_non_gpu_metric("cpu_used_ram"),
-            [non_gpu_data[0][2], non_gpu_data[1][2]],
-        )
-
-    def test_non_gpu_get_metric_value(self):
-        """
-        Test that the non-gpu metric value is correct
-        """
-        self.assertEqual(
-            self.rcm0.get_non_gpu_metric_value("perf_throughput"),
-            sum(
-                [
-                    self.rcm0_non_gpu_metric_values[0]["perf_throughput"],
-                    self.rcm0_non_gpu_metric_values[1]["perf_throughput"],
-                ]
-            ),
-        )
-
-    def test_gpu_get_metric_value(self):
-        """
-        Test that the gpu metric value is correct
-        """
-        self.assertEqual(
-            self.rcm0.get_gpu_metric_value("gpu_used_memory"),
-            self.avg_gpu_metric_values["gpu_used_memory"],
-        )
-
-    def test_get_weighted_non_gpu_metric_value(self):
-        """
-        Test that the non-gpu weighted metric value is correct
-        """
-        sum_weighted_metric_value = (
-            (self.rcm0_non_gpu_metric_values[0]["perf_latency_p99"] * self.weights[0])
-            + (self.rcm0_non_gpu_metric_values[1]["perf_latency_p99"] * self.weights[1])
-        ) / sum(self.weights)
-
-        mean_weighted_metric_value = sum_weighted_metric_value / len(
-            self.rcm0_weighted_non_gpu_metric_values
-        )
-
-        self.assertEqual(
-            self.rcm0.get_weighted_non_gpu_metric_value("perf_latency_p99"),
-            mean_weighted_metric_value,
-        )
-
-    def test_model_specific_pa_params(self):
-        """
-        Test that the model specific PA params are correct
-        """
-        self.assertEqual(
-            self.rcm0.model_specific_pa_params(), self.model_specific_pa_params
-        )
-
-    def test_is_better_than(self):
-        """
-        Test to ensure measurement comparison is working as intended
-        """
-        # RCM0: 1000, 40    RCM1: 500, 30  weights:[1,3]
-        # RCM0-A's throughput is better than RCM1-A (0.5)
-        # RCM0-B's latency is worse than RCM1-B (-0.25)
-        # Factoring in model config weighting
-        # tips this is favor of RCM1 (0.125, -0.1875)
-        self.assertFalse(self.rcm0.is_better_than(self.rcm1))
-
-        # This tips the scale in the favor of RCM0 (0.2, -0.15)
-        self.rcm0.set_model_config_weighting([2, 3])
-        self.assertTrue(self.rcm0.is_better_than(self.rcm1))
-        self.assertGreater(self.rcm0, self.rcm1)
-
-    def test_is_better_than_consistency(self):
-        """
-        Test to ensure measurement comparison is working correctly
-        when the percentage gain is very close
-        """
-        # Throughputs
-        #   RCM2{A/B}: {336,223}   RCM3{A/B}: {270,272}
-        #   RCM2-A is 21.8% better than RCM3-A
-        #   RMC2-B is 19.8% worse than RCM3-B
-        # Therefore, RCM2 is (very slightly) better than RCM3
-        self.assertTrue(self.rcm2.is_better_than(self.rcm3))
-        self.assertFalse(self.rcm3.is_better_than(self.rcm2))
-
-    def test_compare_measurements(self):
-        """
-        Test to ensure compare measurement function returns
-        the correct magnitude
-        # RCM4's throughput is 1000
-        # RCM5's throughput is 2000
-        # Therefore, the magnitude is (RCM5 - RCM4) / avg throughput
-        #                             (2000 - 1000) / 1500
-        """
-        rcm4_vs_rcm5 = self.rcm4.compare_measurements(self.rcm5)
-        self.assertEqual(rcm4_vs_rcm5, 1000 / 1500)
-
-        rcm5_vs_rcm4 = self.rcm5.compare_measurements(self.rcm4)
-        self.assertEqual(rcm5_vs_rcm4, -1000 / 1500)
-
-    def test_is_passing_constraints_none(self):
-        """
-        Test to ensure constraints are reported as passing
-        if none were specified
-        """
-        self.rcm5.set_constraint_manager(
-            construct_constraint_manager(
-                """
-            profile_models:
-              modelA
-            """
-            )
-        )
-        self.assertTrue(self.rcm5.is_passing_constraints())
-
-    def test_is_passing_constraints(self):
-        """
-        Test to ensure constraints are reported as
-        passing/failing if model is above/below
-        throughput threshold
-        """
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 500
-            """
-        )
-        self.rcm5.set_constraint_manager(constraint_manager)
-
-        self.assertTrue(self.rcm5.is_passing_constraints())
-
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 3000
-            """
-        )
-        self.rcm5.set_constraint_manager(constraint_manager)
-
-        self.assertFalse(self.rcm5.is_passing_constraints())
-
-    def test_compare_constraints_none(self):
-        """
-        Checks case where either self or other is passing constraints
-        """
-        # RCM4's throughput is 1000
-        # RCM5's throughput is 2000
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 500
-            """
-        )
-        self.rcm4.set_constraint_manager(constraint_manager)
-
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 2500
-            """
-        )
-        self.rcm5.set_constraint_manager(constraint_manager)
-
-        self.assertEqual(self.rcm4.compare_constraints(self.rcm5), None)
-        self.assertEqual(self.rcm5.compare_constraints(self.rcm4), None)
-
-    def test_compare_constraints_equal(self):
-        """
-        Test to ensure compare constraints reports zero when both
-        RCMs are missing constraints by the same amount
-        """
-        # RCM4's throughput is 1000
-        # RCM5's throughput is 2000
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 1250
-            """
-        )
-        self.rcm4.set_constraint_manager(constraint_manager)
-
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 2500
-            """
-        )
-        self.rcm5.set_constraint_manager(constraint_manager)
-
-        # RCM4 is failing by 20%, RCM5 is failing by 20%
-        self.assertEqual(self.rcm4.compare_constraints(self.rcm5), 0)
-        self.assertEqual(self.rcm5.compare_constraints(self.rcm4), 0)
-
-    def test_compare_constraints_unequal(self):
-        """
-        Test to ensure compare constraints reports the correct
-        value when the RCMs are both failing constraints by different
-        amounts
-        """
-        # RCM4's throughput is 1000
-        # RCM5's throughput is 2000
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 2000
-            """
-        )
-        self.rcm4.set_constraint_manager(constraint_manager)
-
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              modelA:
-                constraints:
-                  perf_throughput:
-                    min: 2500
-            """
-        )
-        self.rcm5.set_constraint_manager(constraint_manager)
-
-        # RCM4 is failing by 50%, RCM5 is failing by 20%
-        self.assertEqual(self.rcm4.compare_constraints(self.rcm5), 0.30)
-        self.assertEqual(self.rcm5.compare_constraints(self.rcm4), -0.30)
-
-    def test_calculate_weighted_percentage_gain(self):
-        """
-        Test to ensure weighted percentage gain is being calculated correctly
-        """
-
-        # RCM0: 1000, 40    RCM1: 500, 30  weights:[1,3]
-        # RCM0-A's throughput is better than RCM1-A (0.5)
-        # RCM0-B's latency is worse than RCM1-B (-0.25)
-        # Factoring in model config weighting
-        # tips this is favor of RCM1 (0.125, -0.1875)
-        # However, by percentage RCM0 will be evaluated as slightly better
-        # 100% on throughput, -25% on latency
-        # Factoring in weighting, RCM0 is slightly better (100 - 75) / 4 = 6.25%
-        self.assertEqual(self.rcm0.calculate_weighted_percentage_gain(self.rcm1), 6.25)
-
-        # Changing the weighting tips the scale in the favor of RCM0 (0.2, -0.15)
-        # And, from a percentage standpoint we get: (200 - 75) / 5 = 25%
-        self.rcm0.set_model_config_weighting([2, 3])
-        self.assertEqual(self.rcm0.calculate_weighted_percentage_gain(self.rcm1), 25.0)
-
-    def test_from_dict(self):
-        """
-        Test to ensure class can be correctly restored from a dictionary
-        """
-        rcm0_json = json.dumps(self.rcm0, default=default_encode)
-
-        rcm0_from_dict = RunConfigMeasurement.from_dict(json.loads(rcm0_json))
-
-        self.assertEqual(
-            rcm0_from_dict.model_variants_name(), self.rcm0.model_variants_name()
-        )
-        self.assertEqual(rcm0_from_dict.gpu_data(), self.rcm0.gpu_data())
-        self.assertEqual(rcm0_from_dict.non_gpu_data(), self.rcm0.non_gpu_data())
-        self.assertEqual(
-            list(rcm0_from_dict.data().values()), list(self.rcm0.data().values())
-        )
-        self.assertEqual(
-            rcm0_from_dict._model_config_measurements,
-            self.rcm0._model_config_measurements,
-        )
-        self.assertEqual(rcm0_from_dict._model_config_weights, [])
-
-    def _construct_rcm0(self):
-        self.model_name = "modelA,modelB"
-        self.model_config_name = ["modelA_config_0", "modelB_config_1"]
-        self.model_variants_name = "".join(self.model_config_name)
-        self.model_specific_pa_params = [
-            {"batch_size": 1, "concurrency": 1},
-            {"batch_size": 2, "concurrency": 2},
-        ]
-
-        self.gpu_metric_values = {
-            "0": {"gpu_used_memory": 6000, "gpu_utilization": 60},
-            "1": {"gpu_used_memory": 10000, "gpu_utilization": 20},
-        }
-        self.avg_gpu_metric_values = {"gpu_used_memory": 8000, "gpu_utilization": 40}
-
-        self.rcm0_non_gpu_metric_values = [
-            {
-                # modelA_config_0
-                "perf_throughput": 1000,
-                "perf_latency_p99": 20,
-                "cpu_used_ram": 1000,
-            },
-            {
-                # modelB_config_1
-                "perf_throughput": 2000,
-                "perf_latency_p99": 40,
-                "cpu_used_ram": 1500,
-            },
-        ]
-
-        self.metric_objectives = [{"perf_throughput": 1}, {"perf_latency_p99": 1}]
-
-        self.weights = [1, 3]
-
-        self.rcm0_weighted_non_gpu_metric_values = []
-        for index, non_gpu_metric_values in enumerate(self.rcm0_non_gpu_metric_values):
-            self.rcm0_weighted_non_gpu_metric_values.append(
-                {
-                    objective: value * self.weights[index] / sum(self.weights)
-                    for (objective, value) in non_gpu_metric_values.items()
-                }
-            )
-
-        self.rcm0 = construct_run_config_measurement(
-            self.model_name,
-            self.model_config_name,
-            self.model_specific_pa_params,
-            self.gpu_metric_values,
-            self.rcm0_non_gpu_metric_values,
-            MagicMock(),
-            self.metric_objectives,
-            self.weights,
-        )
-
-    def _construct_rcm1(self):
-        model_name = "modelA,modelB"
-        model_config_name = ["modelA_config_2", "modelB_config_3"]
-        model_specific_pa_params = [
-            {"batch_size": 3, "concurrency": 3},
-            {"batch_size": 4, "concurrency": 4},
-        ]
-
-        gpu_metric_values = {
-            "0": {"gpu_used_memory": 7000, "gpu_utilization": 40},
-            "1": {"gpu_used_memory": 12000, "gpu_utilization": 30},
-        }
-
-        self.rcm1_non_gpu_metric_values = [
-            {
-                # modelA_config_2
-                "perf_throughput": 500,
-                "perf_latency_p99": 20,
-                "cpu_used_ram": 1000,
-            },
-            {
-                # modelB_config_3
-                "perf_throughput": 1200,
-                "perf_latency_p99": 30,
-                "cpu_used_ram": 1500,
-            },
-        ]
-
-        metric_objectives = [{"perf_throughput": 1}, {"perf_throughput": 1}]
-
-        weights = [1, 3]
-
-        self.rcm1_weighted_non_gpu_metric_values = []
-        for index, non_gpu_metric_values in enumerate(self.rcm1_non_gpu_metric_values):
-            self.rcm1_weighted_non_gpu_metric_values.append(
-                {
-                    objective: value * self.weights[index] / sum(weights)
-                    for (objective, value) in non_gpu_metric_values.items()
-                }
-            )
-
-        self.rcm1 = construct_run_config_measurement(
-            model_name,
-            model_config_name,
-            model_specific_pa_params,
-            gpu_metric_values,
-            self.rcm1_non_gpu_metric_values,
-            MagicMock(),
-            metric_objectives,
-            weights,
-        )
-
-    def _construct_rcm2(self):
-        model_name = "modelA,modelB"
-        model_config_name = ["modelA_config_1", "modelB_config_2"]
-        model_specific_pa_params = [
-            {"batch_size": 3, "concurrency": 3},
-            {"batch_size": 4, "concurrency": 4},
-        ]
-
-        gpu_metric_values = {
-            "0": {"gpu_used_memory": 7000, "gpu_utilization": 40},
-            "1": {"gpu_used_memory": 12000, "gpu_utilization": 30},
-        }
-
-        self.rcm2_non_gpu_metric_values = [
-            {
-                # modelA_config_1
-                "perf_throughput": 336,
-                "perf_latency_p99": 20,
-                "cpu_used_ram": 1000,
-            },
-            {
-                # modelB_config_2
-                "perf_throughput": 223,
-                "perf_latency_p99": 30,
-                "cpu_used_ram": 1500,
-            },
-        ]
-
-        metric_objectives = [{"perf_throughput": 1}, {"perf_throughput": 1}]
-
-        weights = [1, 1]
-
-        self.rcm2_weighted_non_gpu_metric_values = []
-        for index, non_gpu_metric_values in enumerate(self.rcm2_non_gpu_metric_values):
-            self.rcm2_weighted_non_gpu_metric_values.append(
-                {
-                    objective: value * self.weights[index] / sum(weights)
-                    for (objective, value) in non_gpu_metric_values.items()
-                }
-            )
-
-        self.rcm2 = construct_run_config_measurement(
-            model_name,
-            model_config_name,
-            model_specific_pa_params,
-            gpu_metric_values,
-            self.rcm2_non_gpu_metric_values,
-            MagicMock(),
-            metric_objectives,
-            weights,
-        )
-
-    def _construct_rcm3(self):
-        model_name = "modelA,modelB"
-        model_config_name = ["modelA_config_1", "modelB_config_2"]
-        model_specific_pa_params = [
-            {"batch_size": 3, "concurrency": 3},
-            {"batch_size": 4, "concurrency": 4},
-        ]
-
-        gpu_metric_values = {
-            "0": {"gpu_used_memory": 7000, "gpu_utilization": 40},
-            "1": {"gpu_used_memory": 12000, "gpu_utilization": 30},
-        }
-
-        self.rcm3_non_gpu_metric_values = [
-            {
-                # modelA_config_1
-                "perf_throughput": 270,
-                "perf_latency_p99": 20,
-                "cpu_used_ram": 1000,
-            },
-            {
-                # modelB_config_2
-                "perf_throughput": 272,
-                "perf_latency_p99": 30,
-                "cpu_used_ram": 1500,
-            },
-        ]
-
-        metric_objectives = [{"perf_throughput": 1}, {"perf_throughput": 1}]
-
-        weights = [1, 1]
-
-        self.rcm3_weighted_non_gpu_metric_values = []
-        for index, non_gpu_metric_values in enumerate(self.rcm3_non_gpu_metric_values):
-            self.rcm3_weighted_non_gpu_metric_values.append(
-                {
-                    objective: value * self.weights[index] / sum(weights)
-                    for (objective, value) in non_gpu_metric_values.items()
-                }
-            )
-
-        self.rcm3 = construct_run_config_measurement(
-            model_name,
-            model_config_name,
-            model_specific_pa_params,
-            gpu_metric_values,
-            self.rcm3_non_gpu_metric_values,
-            MagicMock(),
-            metric_objectives,
-            weights,
-        )
-
-    def _construct_rcm4(self):
-        model_name = "modelA"
-        model_config_name = ["modelA_config_0"]
-        model_specific_pa_params = [
-            {"batch_size": 1, "concurrency": 1},
-            {"batch_size": 2, "concurrency": 2},
-        ]
-
-        gpu_metric_values = {
-            "0": {"gpu_used_memory": 6000, "gpu_utilization": 60},
-            "1": {"gpu_used_memory": 10000, "gpu_utilization": 20},
-        }
-
-        self.rcm4_non_gpu_metric_values = [
-            {
-                # modelA_config_0
-                "perf_throughput": 1000,
-                "perf_latency_p99": 20,
-                "cpu_used_ram": 1000,
-            },
-        ]
-
-        metric_objectives = [{"perf_throughput": 1}]
-
-        weights = [1]
-
-        self.rcm4_weighted_non_gpu_metric_values = []
-        for index, non_gpu_metric_values in enumerate(self.rcm4_non_gpu_metric_values):
-            self.rcm4_weighted_non_gpu_metric_values.append(
-                {
-                    objective: value * self.weights[index] / sum(self.weights)
-                    for (objective, value) in non_gpu_metric_values.items()
-                }
-            )
-
-        self.rcm4 = construct_run_config_measurement(
-            model_name,
-            model_config_name,
-            model_specific_pa_params,
-            gpu_metric_values,
-            self.rcm4_non_gpu_metric_values,
-            MagicMock(),
-            metric_objectives,
-            weights,
-        )
-
-    def _construct_rcm5(self):
-        model_name = "modelA"
-        model_config_name = ["modelA_config_0"]
-        model_specific_pa_params = [
-            {"batch_size": 1, "concurrency": 1},
-            {"batch_size": 2, "concurrency": 2},
-        ]
-
-        gpu_metric_values = {
-            "0": {"gpu_used_memory": 6000, "gpu_utilization": 60},
-            "1": {"gpu_used_memory": 10000, "gpu_utilization": 20},
-        }
-
-        self.rcm5_non_gpu_metric_values = [
-            {
-                # modelA_config_0
-                "perf_throughput": 2000,
-                "perf_latency_p99": 20,
-                "cpu_used_ram": 1000,
-            },
-        ]
-
-        metric_objectives = [{"perf_throughput": 1}]
-
-        weights = [1]
-
-        self.rcm5_weighted_non_gpu_metric_values = []
-        for index, non_gpu_metric_values in enumerate(self.rcm5_non_gpu_metric_values):
-            self.rcm5_weighted_non_gpu_metric_values.append(
-                {
-                    objective: value * self.weights[index] / sum(self.weights)
-                    for (objective, value) in non_gpu_metric_values.items()
-                }
-            )
-
-        self.rcm5 = construct_run_config_measurement(
-            model_name,
-            model_config_name,
-            model_specific_pa_params,
-            gpu_metric_values,
-            self.rcm5_non_gpu_metric_values,
-            MagicMock(),
-            metric_objectives,
-            weights,
-        )
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_run_config_result_comparator.py b/src/c++/perf_analyzer/tests/test_run_config_result_comparator.py
deleted file mode 100755
index 34a210603..000000000
--- a/src/c++/perf_analyzer/tests/test_run_config_result_comparator.py
+++ /dev/null
@@ -1,258 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from typing import List
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.result.run_config_result_comparator import RunConfigResultComparator
-
-from .common import test_result_collector as trc
-from .common.test_utils import construct_run_config_result
-
-
-class TestRunConfigResultComparatorMethods(trc.TestResultCollector):
-    def setUp(self):
-        self._initialize_metrics()
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_throughput_driven(self):
-        objective_spec = [{"perf_throughput": 2, "perf_latency_p99": 1}]
-        model_weights = [1]
-
-        self._check_run_config_result_comparison(
-            objective_spec=objective_spec,
-            model_weights=model_weights,
-            avg_gpu_metrics1=self.avg_gpu_metrics1,
-            avg_non_gpu_metrics1=self.avg_non_gpu_metrics1,
-            avg_gpu_metrics2=self.avg_gpu_metrics2,
-            avg_non_gpu_metrics2=self.avg_non_gpu_metrics2,
-            expected_result=False,
-        )
-
-    def test_throughput_driven_multi_model(self):
-        objective_spec = [
-            {"perf_throughput": 2, "perf_latency_p99": 1},
-            {"perf_throughput": 2, "perf_latency_p99": 1},
-        ]
-        model_weights = [1, 1]
-
-        self._check_run_config_result_comparison(
-            objective_spec=objective_spec,
-            model_weights=model_weights,
-            avg_gpu_metrics1=self.avg_gpu_metrics1,
-            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_multi1,
-            avg_gpu_metrics2=self.avg_gpu_metrics2,
-            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_multi2,
-            expected_result=False,
-            model_name="test_model",
-            model_config_names=["test_model_config_0", "test_model_config_1"],
-        )
-
-    def test_latency_driven(self):
-        objective_spec = [
-            {"perf_throughput": 1, "perf_latency_p99": 2},
-        ]
-        model_weights = [1]
-
-        self._check_run_config_result_comparison(
-            objective_spec=objective_spec,
-            model_weights=model_weights,
-            avg_gpu_metrics1=self.avg_gpu_metrics1,
-            avg_non_gpu_metrics1=self.avg_non_gpu_metrics1,
-            avg_gpu_metrics2=self.avg_gpu_metrics2,
-            avg_non_gpu_metrics2=self.avg_non_gpu_metrics2,
-            expected_result=True,
-        )
-
-    def test_latency_driven_multi(self):
-        objective_spec = [
-            {"perf_throughput": 1, "perf_latency_p99": 2},
-            {"perf_throughput": 1, "perf_latency_p99": 2},
-        ]
-        model_weights = [1, 1]
-
-        self._check_run_config_result_comparison(
-            objective_spec=objective_spec,
-            model_weights=model_weights,
-            avg_gpu_metrics1=self.avg_gpu_metrics1,
-            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_multi1,
-            avg_gpu_metrics2=self.avg_gpu_metrics2,
-            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_multi2,
-            expected_result=True,
-            model_name="test_model",
-            model_config_names=["test_model_config_0", "test_model_config_1"],
-        )
-
-    def test_equal_weight(self):
-        objective_spec = [{"perf_throughput": 1, "perf_latency_p99": 1}]
-        model_weights = [1]
-
-        self._check_run_config_result_comparison(
-            objective_spec=objective_spec,
-            model_weights=model_weights,
-            avg_gpu_metrics1=self.avg_gpu_metrics1,
-            avg_non_gpu_metrics1=self.avg_non_gpu_metrics1,
-            avg_gpu_metrics2=self.avg_gpu_metrics2,
-            avg_non_gpu_metrics2=self.avg_non_gpu_metrics2,
-            value_step2=2,
-            expected_result=False,
-        )
-
-    def test_equal_weight_multi(self):
-        objective_spec = [
-            {"perf_throughput": 1, "perf_latency_p99": 1},
-            {"perf_throughput": 1, "perf_latency_p99": 1},
-        ]
-        model_weights = [1, 1]
-
-        self._check_run_config_result_comparison(
-            objective_spec=objective_spec,
-            model_weights=model_weights,
-            avg_gpu_metrics1=self.avg_gpu_metrics1,
-            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_multi1,
-            avg_gpu_metrics2=self.avg_gpu_metrics2,
-            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_multi2,
-            value_step2=2,
-            expected_result=False,
-            model_name="test_model",
-            model_config_names=["test_model_config_0", "test_model_config_1"],
-        )
-
-    def test_unequal_weight_multi(self):
-        """
-        Tests that changing just the model weighting will result in a different
-        config being selected as better
-        """
-        objective_spec = [
-            {"perf_throughput": 1, "perf_latency_p99": 1},
-            {"perf_throughput": 1, "perf_latency_p99": 1},
-        ]
-
-        # With equal weighting - config 1 is better
-        model_weights = [1, 1]
-        self._check_run_config_result_comparison(
-            objective_spec=objective_spec,
-            model_weights=model_weights,
-            avg_gpu_metrics1=self.avg_gpu_metrics1,
-            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_weighted_multi1,
-            avg_gpu_metrics2=self.avg_gpu_metrics2,
-            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_weighted_multi2,
-            value_step2=2,
-            expected_result=False,
-            model_name="test_model",
-            model_config_names=["test_model_config_0", "test_model_config_1"],
-        )
-
-        # With an unequal weighting - config 0 is better
-        model_weights = [3, 1]
-        self._check_run_config_result_comparison(
-            objective_spec=objective_spec,
-            model_weights=model_weights,
-            avg_gpu_metrics1=self.avg_gpu_metrics1,
-            avg_non_gpu_metrics1=self.avg_non_gpu_metrics_weighted_multi1,
-            avg_gpu_metrics2=self.avg_gpu_metrics2,
-            avg_non_gpu_metrics2=self.avg_non_gpu_metrics_weighted_multi2,
-            value_step2=2,
-            expected_result=True,
-            model_name="test_model",
-            model_config_names=["test_model_config_0", "test_model_config_1"],
-        )
-
-    def _check_run_config_result_comparison(
-        self,
-        objective_spec,
-        model_weights: List[int],
-        avg_gpu_metrics1,
-        avg_gpu_metrics2,
-        avg_non_gpu_metrics1,
-        avg_non_gpu_metrics2,
-        value_step1=1,
-        value_step2=1,
-        expected_result=0,
-        model_name="test_model",
-        model_config_names=["test_model"],
-    ):
-        """
-        Helper function that takes all the data needed to
-        construct two RunConfigResults, constructs and runs a
-        comparator and checks that it produces the expected
-        value.
-        """
-
-        result_comparator = RunConfigResultComparator(
-            metric_objectives_list=objective_spec, model_weights=model_weights
-        )
-
-        result1 = construct_run_config_result(
-            avg_gpu_metric_values=avg_gpu_metrics1,
-            avg_non_gpu_metric_values_list=avg_non_gpu_metrics1,
-            comparator=result_comparator,
-            value_step=value_step1,
-            run_config=MagicMock(),
-            constraint_manager=MagicMock(),
-            model_name=model_name,
-            model_config_names=model_config_names,
-        )
-
-        result2 = construct_run_config_result(
-            avg_gpu_metric_values=avg_gpu_metrics2,
-            avg_non_gpu_metric_values_list=avg_non_gpu_metrics2,
-            comparator=result_comparator,
-            value_step=value_step2,
-            run_config=MagicMock(),
-            constraint_manager=MagicMock(),
-            model_name=model_name,
-            model_config_names=model_config_names,
-        )
-
-        self.assertEqual(
-            result_comparator.is_better_than(result1, result2), expected_result
-        )
-
-    def _initialize_metrics(self):
-        self.avg_gpu_metrics1 = {0: {"gpu_used_memory": 5000, "gpu_utilization": 50}}
-        self.avg_gpu_metrics2 = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-
-        self.avg_non_gpu_metrics1 = [{"perf_throughput": 100, "perf_latency_p99": 4000}]
-
-        self.avg_non_gpu_metrics2 = [{"perf_throughput": 200, "perf_latency_p99": 8000}]
-
-        self.avg_non_gpu_metrics_multi1 = [
-            {"perf_throughput": 100, "perf_latency_p99": 4000},
-            {"perf_throughput": 150, "perf_latency_p99": 6000},
-        ]
-
-        self.avg_non_gpu_metrics_multi2 = [
-            {"perf_throughput": 200, "perf_latency_p99": 8000},
-            {"perf_throughput": 250, "perf_latency_p99": 10000},
-        ]
-
-        self.avg_non_gpu_metrics_weighted_multi1 = [
-            {"perf_throughput": 1000, "perf_latency_p99": 50},
-            {"perf_throughput": 1000, "perf_latency_p99": 100},
-        ]
-
-        self.avg_non_gpu_metrics_weighted_multi2 = [
-            {"perf_throughput": 2000, "perf_latency_p99": 200},
-            {"perf_throughput": 500, "perf_latency_p99": 20},
-        ]
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_search_config.py b/src/c++/perf_analyzer/tests/test_search_config.py
deleted file mode 100755
index b599ffb70..000000000
--- a/src/c++/perf_analyzer/tests/test_search_config.py
+++ /dev/null
@@ -1,88 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from model_analyzer.config.generate.search_config import SearchConfig
-from model_analyzer.config.generate.search_dimension import SearchDimension
-from model_analyzer.config.generate.search_dimensions import SearchDimensions
-
-from .common import test_result_collector as trc
-
-
-class TestSearchConfig(trc.TestResultCollector):
-    def test_basic(self):
-        sc = SearchConfig(SearchDimensions(), 0, 0)
-        self.assertEqual(0, sc.get_num_dimensions())
-
-    def test_config(self):
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        sc = SearchConfig(dimensions=dims, radius=4, min_initialized=2)
-
-        self.assertEqual(2, sc.get_num_dimensions())
-        self.assertEqual(4, sc.get_radius())
-        self.assertEqual(2, sc.get_min_initialized())
-
-        self.assertEqual("foo", sc.get_dimension(0).get_name())
-        self.assertEqual("bar", sc.get_dimension(1).get_name())
-
-        self.assertEqual(7, sc.get_dimension(0).get_value_at_idx(6))
-        self.assertEqual(64, sc.get_dimension(1).get_value_at_idx(6))
-
-    def test_get_min_indexes(self):
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR, 1, 10),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-        sc = SearchConfig(dims, 0, 0)
-
-        self.assertEqual([1, 0], sc.get_min_indexes())
-
-    def test_get_neighborhood_config(self):
-        """
-        Test that we can get a NeighborhoodConfig from a SearchConfig and properly override the radius
-        """
-        dims = SearchDimensions()
-        dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_EXPONENTIAL),
-            ],
-        )
-
-        sc = SearchConfig(dimensions=dims, radius=4, min_initialized=2)
-
-        nc = sc.get_neighborhood_config(radius=5)
-        self.assertEqual(2, nc.get_num_dimensions())
-        self.assertEqual(5, nc.get_radius())
-        self.assertEqual(2, nc.get_min_initialized())
-
-        self.assertEqual("foo", nc.get_dimension(0).get_name())
-        self.assertEqual("bar", nc.get_dimension(1).get_name())
-
-        self.assertEqual(7, nc.get_dimension(0).get_value_at_idx(6))
-        self.assertEqual(64, nc.get_dimension(1).get_value_at_idx(6))
diff --git a/src/c++/perf_analyzer/tests/test_search_dimension.py b/src/c++/perf_analyzer/tests/test_search_dimension.py
deleted file mode 100755
index 1f51b2ea6..000000000
--- a/src/c++/perf_analyzer/tests/test_search_dimension.py
+++ /dev/null
@@ -1,53 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from model_analyzer.config.generate.search_dimension import SearchDimension
-
-from .common import test_result_collector as trc
-
-
-class TestSearchDimension(trc.TestResultCollector):
-    def test_linear(self):
-        sd = SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR, 0, 10)
-
-        self.assertEqual("foo", sd.get_name())
-        self.assertEqual(0, sd.get_min_idx())
-        self.assertEqual(10, sd.get_max_idx())
-        self.assertEqual(1, sd.get_value_at_idx(0))
-        self.assertEqual(2, sd.get_value_at_idx(1))
-        self.assertEqual(3, sd.get_value_at_idx(2))
-
-    def test_exponential(self):
-        sd = SearchDimension("foo", SearchDimension.DIMENSION_TYPE_EXPONENTIAL)
-        self.assertEqual(1, sd.get_value_at_idx(0))
-        self.assertEqual(2, sd.get_value_at_idx(1))
-        self.assertEqual(4, sd.get_value_at_idx(2))
-        self.assertEqual(8, sd.get_value_at_idx(3))
-
-    def test_out_of_bounds(self):
-        sd = SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR, 2, 10)
-
-        with self.assertRaises(IndexError):
-            sd.get_value_at_idx(1)
-
-        with self.assertRaises(IndexError):
-            sd.get_value_at_idx(11)
-
-    def test_no_max(self):
-        sd = SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR)
-
-        # Confirm no assert
-        sd.get_value_at_idx(100000000)
diff --git a/src/c++/perf_analyzer/tests/test_search_dimensions.py b/src/c++/perf_analyzer/tests/test_search_dimensions.py
deleted file mode 100755
index e9a0e79ad..000000000
--- a/src/c++/perf_analyzer/tests/test_search_dimensions.py
+++ /dev/null
@@ -1,79 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from model_analyzer.config.generate.coordinate import Coordinate
-from model_analyzer.config.generate.search_dimension import SearchDimension
-from model_analyzer.config.generate.search_dimensions import SearchDimensions
-
-from .common import test_result_collector as trc
-
-
-class TestSearchDimensions(trc.TestResultCollector):
-    def setUp(self):
-        self._dims = SearchDimensions()
-        self._dims.add_dimensions(
-            0,
-            [
-                SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR),
-                SearchDimension("bar", SearchDimension.DIMENSION_TYPE_LINEAR),
-            ],
-        )
-
-        self._dims.add_dimensions(
-            1, [SearchDimension("foo", SearchDimension.DIMENSION_TYPE_LINEAR)]
-        )
-
-    def test_list_behavior(self):
-        """
-        Confirm the following list-like behavior:
-            - get length like a list
-            - get from index like a list
-            - iterate over values
-            - raise IndexError on out of bounds index
-        """
-        dims = self._dims
-        self.assertEqual(len(dims), 3)
-
-        dim0 = dims[0]
-        dim1 = dims[1]
-        dim2 = dims[2]
-        self.assertEqual(dim0.get_name(), "foo")
-        self.assertEqual(dim1.get_name(), "bar")
-        self.assertEqual(dim2.get_name(), "foo")
-
-        count = 0
-        for dim in dims:
-            self.assertTrue(isinstance(dim, SearchDimension))
-            count += 1
-        self.assertEqual(count, 3)
-
-        with self.assertRaises(IndexError):
-            _ = dims[3]
-
-    def test_get_values(self):
-        """
-        Test get_values_for_coordinate() functionality
-        """
-        dims = self._dims
-        vals = dims.get_values_for_coordinate(Coordinate([0, 2, 4]))
-        expected_vals = {
-            0: {
-                "foo": 1,
-                "bar": 3,
-            },
-            1: {"foo": 5},
-        }
-        self.assertEqual(vals, expected_vals)
diff --git a/src/c++/perf_analyzer/tests/test_sorted_results.py b/src/c++/perf_analyzer/tests/test_sorted_results.py
deleted file mode 100755
index 1b8c76715..000000000
--- a/src/c++/perf_analyzer/tests/test_sorted_results.py
+++ /dev/null
@@ -1,306 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from random import sample
-from unittest.mock import MagicMock, patch
-
-import yaml
-
-from model_analyzer.result.run_config_result_comparator import RunConfigResultComparator
-from model_analyzer.result.sorted_results import SortedResults
-
-from .common import test_result_collector as trc
-from .common.test_utils import (
-    construct_constraint_manager,
-    construct_run_config,
-    construct_run_config_result,
-)
-
-
-class TestSortedResults(trc.TestResultCollector):
-    def setUp(self):
-        objective_spec = {"perf_throughput": 10, "perf_latency_p99": 5}
-        self.sorted_results = SortedResults()
-        self.result_comparator = RunConfigResultComparator(
-            metric_objectives_list=[objective_spec], model_weights=[1]
-        )
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_add_results(self):
-        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-        avg_non_gpu_metrics = {"perf_throughput": 100, "perf_latency_p99": 4000}
-
-        for _ in range(10):
-            self.sorted_results.add_result(
-                construct_run_config_result(
-                    avg_gpu_metric_values=avg_gpu_metrics,
-                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
-                    constraint_manager=MagicMock(),
-                    comparator=self.result_comparator,
-                )
-            )
-
-        results = self.sorted_results.results()
-        self.assertEqual(len(results), 10)
-
-    def test_add_results_same_config(self):
-        """
-        Test that additional results with the same model config variant
-        will be added to the same entry in the list
-        """
-        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-        avg_non_gpu_metrics = {"perf_throughput": 100, "perf_latency_p99": 4000}
-
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              test_model
-            """
-        )
-        for _ in range(10):
-            run_config = construct_run_config("modelA", "model_config_0", "key_A")
-            self.sorted_results.add_result(
-                construct_run_config_result(
-                    avg_gpu_metric_values=avg_gpu_metrics,
-                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
-                    comparator=self.result_comparator,
-                    constraint_manager=constraint_manager,
-                    run_config=run_config,
-                )
-            )
-
-        results = self.sorted_results.results()
-        self.assertEqual(len(results), 1)
-
-    def test_add_results_dynamic_sorting(self):
-        """
-        Tests the case where configs A & B where A is better than B
-        Then add a new measurement to B, which now makes B better than A
-        """
-        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-
-        run_config_A = construct_run_config("model", "model_config_A", "key_A")
-        run_config_B = construct_run_config("model", "model_config_B", "key_B")
-        run_config_list = [run_config_A, run_config_B]
-
-        constraint_manager = construct_constraint_manager(
-            """
-            profile_models:
-              model
-            """
-        )
-
-        for i in sample(range(2), 2):
-            avg_non_gpu_metrics = {
-                "perf_throughput": 100 - 10 * i,
-                "perf_latency_p99": 4000,
-            }
-            self.sorted_results.add_result(
-                construct_run_config_result(
-                    avg_gpu_metric_values=avg_gpu_metrics,
-                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
-                    comparator=self.result_comparator,
-                    model_name="model",
-                    model_config_names=["model_config_0"],
-                    constraint_manager=constraint_manager,
-                    run_config=run_config_list[i],
-                )
-            )
-
-        all_results = self.sorted_results.top_n_results(n=SortedResults.GET_ALL_RESULTS)
-
-        self.assertEqual(
-            all_results[0].run_config().model_variants_name(), "model_config_A"
-        )
-        self.assertEqual(
-            all_results[1].run_config().model_variants_name(), "model_config_B"
-        )
-
-        avg_non_gpu_metrics = {"perf_throughput": 200, "perf_latency_p99": 4000}
-
-        self.sorted_results.add_result(
-            construct_run_config_result(
-                avg_gpu_metric_values=avg_gpu_metrics,
-                avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
-                comparator=self.result_comparator,
-                model_name="model",
-                model_config_names=["model_config_0"],
-                constraint_manager=constraint_manager,
-                run_config=run_config_B,
-            )
-        )
-
-        all_results = self.sorted_results.top_n_results(n=SortedResults.GET_ALL_RESULTS)
-
-        self.assertEqual(
-            all_results[0].run_config().model_variants_name(), "model_config_B"
-        )
-        self.assertEqual(
-            all_results[1].run_config().model_variants_name(), "model_config_A"
-        )
-
-    def test_add_results_all_failing(self):
-        """
-        Test the case where we have only failing results
-        """
-        # Create constraint_manager for all model names
-        constraint_manager = construct_constraint_manager(
-            yaml.dump(
-                {
-                    "profile_models": {
-                        str(i): {"constraints": {"perf_throughput": {"min": 1000}}}
-                        for i in range(10)
-                    }
-                }
-            )
-        )
-
-        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-        for i in sample(range(10), 10):
-            model_name = str(i)
-            model_config_names = [f"{model_name}_config_0"]
-
-            avg_non_gpu_metrics = {
-                "perf_throughput": 100 + 10 * i,
-                "perf_latency_p99": 4000,
-            }
-            self.sorted_results.add_result(
-                construct_run_config_result(
-                    avg_gpu_metric_values=avg_gpu_metrics,
-                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
-                    comparator=self.result_comparator,
-                    constraint_manager=constraint_manager,
-                    model_config_names=model_config_names,
-                    model_name=model_name,
-                )
-            )
-
-        top_5_results = self.sorted_results.top_n_results(n=5)
-        self.assertEqual(top_5_results[0].model_name(), "9")
-        self.assertEqual(top_5_results[1].model_name(), "8")
-        self.assertEqual(top_5_results[2].model_name(), "7")
-        self.assertEqual(top_5_results[3].model_name(), "6")
-        self.assertEqual(top_5_results[4].model_name(), "5")
-
-    def test_add_results_failing_to_passing(self):
-        """
-        Test the case where we have only failing results
-        and then a measurement makes one of the results passing
-        """
-
-        # Create constraint_manager for all model names
-        constraint_manager = construct_constraint_manager(
-            yaml.dump(
-                {
-                    "profile_models": {
-                        str(i): {"constraints": {"perf_throughput": {"min": 1000}}}
-                        for i in sample(range(10), 10)
-                    }
-                }
-            )
-        )
-
-        # Create 10 failing results
-        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-        for i in sample(range(10), 10):
-            model_name = str(i)
-            model_config_names = [f"{model_name}_config_0"]
-
-            avg_non_gpu_metrics = {
-                "perf_throughput": 100 + 10 * i,
-                "perf_latency_p99": 4000,
-            }
-            self.sorted_results.add_result(
-                construct_run_config_result(
-                    avg_gpu_metric_values=avg_gpu_metrics,
-                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
-                    comparator=self.result_comparator,
-                    constraint_manager=constraint_manager,
-                    model_config_names=model_config_names,
-                    model_name=model_name,
-                )
-            )
-
-        # Now add a measurement to the last result so that it now passes the constraint
-        avg_non_gpu_metrics = {"perf_throughput": 2000, "perf_latency_p99": 4000}
-
-        model_name = "9"
-        model_config_names = [f"{model_name}_config_0"]
-        self.sorted_results.add_result(
-            construct_run_config_result(
-                avg_gpu_metric_values=avg_gpu_metrics,
-                avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
-                comparator=self.result_comparator,
-                constraint_manager=constraint_manager,
-                model_config_names=model_config_names,
-                model_name=model_name,
-            )
-        )
-
-        top_5_results = self.sorted_results.top_n_results(n=5)
-        self.assertEqual(len(top_5_results), 1)
-        self.assertEqual(top_5_results[0].model_name(), "9")
-
-    def test_top_n_results(self):
-        avg_gpu_metrics = {0: {"gpu_used_memory": 6000, "gpu_utilization": 60}}
-        constraint_manager = construct_constraint_manager(
-            yaml.dump({"profile_models": {str(i): {} for i in range(10)}})
-        )
-
-        for i in sample(range(10), 10):
-            model_name = str(i)
-            model_config_names = [f"{model_name}_config_0"]
-
-            avg_non_gpu_metrics = {
-                "perf_throughput": 100 + 10 * i,
-                "perf_latency_p99": 4000,
-            }
-            self.sorted_results.add_result(
-                construct_run_config_result(
-                    avg_gpu_metric_values=avg_gpu_metrics,
-                    avg_non_gpu_metric_values_list=[avg_non_gpu_metrics],
-                    comparator=self.result_comparator,
-                    model_name=model_name,
-                    model_config_names=model_config_names,
-                    constraint_manager=constraint_manager,
-                )
-            )
-
-        top_5_results = self.sorted_results.top_n_results(n=5)
-        self.assertEqual(top_5_results[0].model_name(), "9")
-        self.assertEqual(top_5_results[1].model_name(), "8")
-        self.assertEqual(top_5_results[2].model_name(), "7")
-        self.assertEqual(top_5_results[3].model_name(), "6")
-        self.assertEqual(top_5_results[4].model_name(), "5")
-
-        all_results = self.sorted_results.top_n_results(n=SortedResults.GET_ALL_RESULTS)
-        self.assertEqual(all_results[0].model_name(), "9")
-        self.assertEqual(all_results[1].model_name(), "8")
-        self.assertEqual(all_results[2].model_name(), "7")
-        self.assertEqual(all_results[3].model_name(), "6")
-        self.assertEqual(all_results[4].model_name(), "5")
-        self.assertEqual(all_results[5].model_name(), "4")
-        self.assertEqual(all_results[6].model_name(), "3")
-        self.assertEqual(all_results[7].model_name(), "2")
-        self.assertEqual(all_results[8].model_name(), "1")
-        self.assertEqual(all_results[9].model_name(), "0")
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_state_manager.py b/src/c++/perf_analyzer/tests/test_state_manager.py
deleted file mode 100755
index 2a5898210..000000000
--- a/src/c++/perf_analyzer/tests/test_state_manager.py
+++ /dev/null
@@ -1,148 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.state.analyzer_state_manager import AnalyzerStateManager
-
-from .common import test_result_collector as trc
-from .common.test_utils import evaluate_mock_config
-from .mocks.mock_glob import MockGlobMethods
-from .mocks.mock_io import MockIOMethods
-from .mocks.mock_json import MockJSONMethods
-from .mocks.mock_os import MockOSMethods
-
-
-class TestAnalyzerStateManagerMethods(trc.TestResultCollector):
-    @patch(
-        "model_analyzer.config.input.config_command_profile.ConfigCommandProfile._preprocess_and_verify_arguments",
-        MagicMock(),
-    )
-    def setUp(self):
-        args = [
-            "model-analyzer",
-            "profile",
-            "--model-repository",
-            "cli_repository",
-            "-f",
-            "path-to-config-file",
-            "--profile-models",
-            "test_model",
-        ]
-        yaml_str = """
-            export_path: /test_export_path/
-        """
-
-        # start mocks
-        self.mock_io = MockIOMethods(
-            mock_paths=["model_analyzer.state.analyzer_state_manager"]
-        )
-        self.mock_json = MockJSONMethods()
-        self.mock_os = MockOSMethods(
-            mock_paths=[
-                "model_analyzer.state.analyzer_state_manager",
-                "model_analyzer.config.input.config_utils",
-            ]
-        )
-        self.mock_glob = MockGlobMethods()
-
-        self.mock_io.start()
-        self.mock_json.start()
-        self.mock_os.start()
-        self.mock_glob.start()
-
-        config = evaluate_mock_config(args, yaml_str, subcommand="profile")
-
-        # state manager
-        self.state_manager = AnalyzerStateManager(config=config, server=None)
-
-    def tearDown(self):
-        patch.stopall()
-
-    def test_set_get_state_variables(self):
-        self.mock_os.set_os_path_exists_return_value(False)
-        self.state_manager.load_checkpoint(checkpoint_required=False)
-
-        vars = [f"test_var{j}" for j in range(10)]
-        for i, name in enumerate(vars):
-            self.state_manager.set_state_variable(name, i)
-
-        for i, name in enumerate(vars):
-            self.assertEqual(self.state_manager.get_state_variable(name), i)
-
-        for i, name in enumerate(vars):
-            self.state_manager.set_state_variable(name, 9 - i)
-
-        for i, name in enumerate(vars):
-            self.assertEqual(self.state_manager.get_state_variable(name), 9 - i)
-
-    def test_load_checkpoint(self):
-        # Load checkpoint without ckpt files
-        self.mock_os.set_os_path_exists_return_value(False)
-        self.state_manager.load_checkpoint(checkpoint_required=False)
-        self.assertTrue(self.state_manager.starting_fresh_run())
-
-        # Load checkpoint files with ckpt files
-        self.mock_os.set_os_path_exists_return_value(True)
-        self.mock_os.set_os_path_join_return_value("0.ckpt")
-        self.state_manager.load_checkpoint(checkpoint_required=False)
-        self.assertFalse(self.state_manager.starting_fresh_run())
-
-        # Load checkpoint throws error
-        self.mock_json.set_json_load_side_effect(EOFError)
-        with self.assertRaises(
-            TritonModelAnalyzerException,
-            msg="Checkpoint file 0.ckpt is"
-            " empty or corrupted. Remove it from checkpoint"
-            " directory.",
-        ):
-            self.mock_os.set_os_path_exists_return_value(True)
-            self.mock_os.set_os_path_join_return_value("0.ckpt")
-            self.state_manager.load_checkpoint(checkpoint_required=False)
-            self.assertFalse(self.state_manager.starting_fresh_run())
-
-    def test_latest_checkpoint(self):
-        # No checkpoints
-        self.mock_glob.set_glob_return_value([])
-        self.assertEqual(self.state_manager._latest_checkpoint(), -1)
-
-        # single checkpoint file
-        for i in range(5):
-            self.mock_glob.set_glob_return_value([f"{i}.ckpt"])
-            self.assertEqual(self.state_manager._latest_checkpoint(), i)
-
-        # Multiple checkpoint files consecutive, sorted
-        self.mock_glob.set_glob_return_value([f"{i}.ckpt" for i in range(5)])
-        self.assertEqual(self.state_manager._latest_checkpoint(), 4)
-
-        # Multiple checkpoint files consecutive, unsorted
-        self.mock_glob.set_glob_return_value([f"{i}.ckpt" for i in range(5, 1, -1)])
-        self.assertEqual(self.state_manager._latest_checkpoint(), 5)
-
-        # Multiple files nonconsecutive unsorted
-        self.mock_glob.set_glob_return_value([f"{i}.ckpt" for i in [1, 3, 5, 2, 0, 4]])
-        self.assertEqual(self.state_manager._latest_checkpoint(), 5)
-
-        # Malformed checkpoint filename
-        self.mock_glob.set_glob_return_value(["XYZ.ckpt"])
-        with self.assertRaises(TritonModelAnalyzerException):
-            self.state_manager._latest_checkpoint()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_triton_client.py b/src/c++/perf_analyzer/tests/test_triton_client.py
deleted file mode 100755
index f63623e5e..000000000
--- a/src/c++/perf_analyzer/tests/test_triton_client.py
+++ /dev/null
@@ -1,214 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os
-import tempfile
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.device.gpu_device import GPUDevice
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.triton.client.client_factory import TritonClientFactory
-from model_analyzer.triton.server.server_config import TritonServerConfig
-from model_analyzer.triton.server.server_factory import TritonServerFactory
-
-from .common import test_result_collector as trc
-from .mocks.mock_client import MockTritonClientMethods
-from .mocks.mock_server_docker import MockServerDockerMethods
-
-# Test parameters
-MODEL_REPOSITORY_PATH = "test_repo"
-TRITON_IMAGE = "test_image"
-CONFIG_TEST_ARG = "url"
-GRPC_URL = "test_grpc_url"
-HTTP_URL = "test_http_url"
-TEST_MODEL_NAME = "test_model"
-
-
-class TestTritonClientMethods(trc.TestResultCollector):
-    def setUp(self):
-        # GPUs
-        gpus = [GPUDevice("TEST_DEVICE_NAME", 0, "TEST_PCI_BUS_ID", "TEST_UUID")]
-
-        # Mocks
-        self.server_docker_mock = MockServerDockerMethods()
-        self.tritonclient_mock = MockTritonClientMethods()
-        self.server_docker_mock.start()
-        self.tritonclient_mock.start()
-
-        # Create server config
-        self.server_config = TritonServerConfig()
-        self.server_config["model-repository"] = MODEL_REPOSITORY_PATH
-        self.server_config["model-control-mode"] = "explicit"
-
-        # Set CUDA_VISIBLE_DEVICES
-        os.environ["CUDA_VISIBLE_DEVICES"] = "0"
-
-        # Create and start the server
-        self.server = TritonServerFactory.create_server_docker(
-            image=TRITON_IMAGE, config=self.server_config, gpus=gpus
-        )
-
-    def test_create_client(self):
-        # Create GRPC client
-        TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
-        self.tritonclient_mock.assert_created_grpc_client_with_args(url=GRPC_URL)
-
-        # Create HTTP client
-        TritonClientFactory.create_http_client(server_url=HTTP_URL)
-        self.tritonclient_mock.assert_created_http_client_with_args(url=HTTP_URL)
-
-    def test_wait_for_server_ready(self):
-        # For reuse
-        def _test_with_client(self, client):
-            with self.assertRaises(
-                TritonModelAnalyzerException,
-                msg="Expected Exception trying" " wait for server ready",
-            ):
-                self.tritonclient_mock.raise_exception_on_wait_for_server_ready()
-                client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
-            self.tritonclient_mock.reset()
-
-            with self.assertRaises(
-                TritonModelAnalyzerException,
-                msg="Expected Exception on" " server not ready",
-            ):
-                self.tritonclient_mock.set_server_not_ready()
-                client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
-
-            self.tritonclient_mock.reset()
-            client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
-
-        # HTTP client
-        client = TritonClientFactory.create_http_client(server_url=HTTP_URL)
-        self.tritonclient_mock.assert_created_http_client_with_args(HTTP_URL)
-        _test_with_client(self, client)
-        self.tritonclient_mock.assert_http_client_waited_for_server_ready()
-
-        # GRPC client
-        client = TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
-        self.tritonclient_mock.assert_created_grpc_client_with_args(GRPC_URL)
-        _test_with_client(self, client)
-        self.tritonclient_mock.assert_grpc_client_waited_for_server_ready()
-
-    def test_wait_for_server_ready_with_invalid_argument(self):
-        """
-        Tests that we detect when an invalid argument is passed to the server
-        """
-        log_file = tempfile.NamedTemporaryFile()
-        log_file.write(b"Unexpected argument: UNKNOWN_CMD")
-
-        client = TritonClientFactory.create_http_client(server_url=HTTP_URL)
-        self.tritonclient_mock.raise_exception_on_wait_for_server_ready()
-
-        with self.assertRaises(TritonModelAnalyzerException):
-            client.wait_for_server_ready(
-                num_retries=1, sleep_time=0.1, log_file=log_file
-            )
-
-    def test_wait_for_model_ready(self):
-        # For reuse
-        def _test_with_client(self, client):
-            client.wait_for_model_ready(
-                model_name=TEST_MODEL_NAME, num_retries=1, sleep_time=0.1
-            )
-            self.tritonclient_mock.reset()
-
-            self.tritonclient_mock.set_model_not_ready()
-            self.assertTrue(
-                client.wait_for_model_ready(
-                    model_name=TEST_MODEL_NAME, num_retries=2, sleep_time=0.1
-                ),
-                -1,
-            )
-            self.tritonclient_mock.reset()
-            client.wait_for_model_ready(
-                model_name=TEST_MODEL_NAME, num_retries=1, sleep_time=0.1
-            )
-
-        # HTTP client
-        client = TritonClientFactory.create_http_client(server_url=HTTP_URL)
-        self.tritonclient_mock.assert_created_http_client_with_args(HTTP_URL)
-        _test_with_client(self, client)
-        self.tritonclient_mock.assert_http_client_waited_for_model_ready(
-            model_name=TEST_MODEL_NAME
-        )
-
-        # GRPC client
-        client = TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
-        self.tritonclient_mock.assert_created_grpc_client_with_args(GRPC_URL)
-        _test_with_client(self, client)
-        self.tritonclient_mock.assert_grpc_client_waited_for_model_ready(
-            model_name=TEST_MODEL_NAME
-        )
-
-    def test_load_unload_model(self):
-        # Create client
-        client = TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
-        self.tritonclient_mock.assert_created_grpc_client_with_args(GRPC_URL)
-
-        # Start the server and wait till it is ready
-        self.server.start()
-        client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
-
-        # Try to load a dummy model and expect error
-        self.tritonclient_mock.raise_exception_on_load()
-        self.assertTrue(client.load_model("dummy"), -1)
-
-        self.tritonclient_mock.reset()
-
-        # Load the test model
-        client.load_model(TEST_MODEL_NAME)
-        client.wait_for_model_ready(TEST_MODEL_NAME, num_retries=1, sleep_time=0.1)
-
-        # Try to unload a dummy model and expect error
-        self.tritonclient_mock.raise_exception_on_unload()
-        self.assertTrue(client.unload_model("dummy"), -1)
-
-        self.tritonclient_mock.reset()
-
-        # Unload the test model
-        client.unload_model(TEST_MODEL_NAME)
-
-        self.server.stop()
-
-    def test_get_model_config(self):
-        # Create client
-        client = TritonClientFactory.create_grpc_client(server_url=GRPC_URL)
-        self.tritonclient_mock.assert_created_grpc_client_with_args(GRPC_URL)
-
-        # Start the server and wait till it is ready
-        self.server.start()
-        client.wait_for_server_ready(num_retries=1, sleep_time=0.1)
-
-        # Set model config, and try to get it
-        test_model_config_dict = {"config": "test_config"}
-        self.tritonclient_mock.set_model_config(test_model_config_dict)
-        self.assertEqual(
-            client.get_model_config(TEST_MODEL_NAME, num_retries=1), "test_config"
-        )
-
-    def tearDown(self):
-        # In case test raises exception
-        if self.server is not None:
-            self.server.stop()
-        self.server_docker_mock.stop()
-        self.tritonclient_mock.stop()
-        patch.stopall()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_triton_server.py b/src/c++/perf_analyzer/tests/test_triton_server.py
deleted file mode 100755
index b083b7567..000000000
--- a/src/c++/perf_analyzer/tests/test_triton_server.py
+++ /dev/null
@@ -1,361 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2020-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import patch
-
-from model_analyzer.device.gpu_device import GPUDevice
-from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
-from model_analyzer.triton.server.server_config import TritonServerConfig
-from model_analyzer.triton.server.server_factory import TritonServerFactory
-
-from .common import test_result_collector as trc
-from .mocks.mock_gpu_device_factory import MockGPUDeviceFactory
-from .mocks.mock_os import MockOSMethods
-from .mocks.mock_server_docker import MockServerDockerMethods
-from .mocks.mock_server_local import MockServerLocalMethods
-
-# Test parameters
-MODEL_REPOSITORY_PATH = "test_repo"
-TRITON_LOCAL_BIN_PATH = "test_bin_path/tritonserver"
-TRITON_DOCKER_BIN_PATH = "tritonserver"
-TRITON_IMAGE = "test_image"
-CONFIG_TEST_ARG = "exit-on-error"
-CLI_TO_STRING_TEST_ARGS = {
-    "allow-grpc": True,
-    "min-supported-compute-capability": 7.5,
-    "metrics-port": 8000,
-    "model-repository": MODEL_REPOSITORY_PATH,
-}
-
-
-class TestTritonServerMethods(trc.TestResultCollector):
-    def setUp(self):
-        # GPUs for this test
-        self._sys_gpus = [
-            GPUDevice("TEST_DEVICE_NAME", 0, "TEST_BUS_ID0", "TEST_UUID0"),
-            GPUDevice("TEST_DEVICE_NAME", 1, "TEST_BUS_ID1", "TEST_UUID1"),
-        ]
-
-        # Mock
-        self.gpu_device_factory_mock = MockGPUDeviceFactory(self._sys_gpus)
-        self.server_docker_mock = MockServerDockerMethods()
-        self.server_local_mock = MockServerLocalMethods()
-        self.os_mock = MockOSMethods(
-            mock_paths=[
-                "model_analyzer.triton.server.server_local",
-                "tests.mocks.mock_server_local",
-            ]
-        )
-        self.gpu_device_factory_mock.start()
-        self.server_docker_mock.start()
-        self.server_local_mock.start()
-        self.os_mock.start()
-
-        # server setup
-        self.server = None
-
-    def test_server_config(self):
-        # Create a TritonServerConfig
-        server_config = TritonServerConfig()
-        server_config["model-repository"] = MODEL_REPOSITORY_PATH
-
-        # Check config initializations
-        self.assertIsNone(
-            server_config[CONFIG_TEST_ARG],
-            msg="Server config had unexpected initial" f"value for {CONFIG_TEST_ARG}",
-        )
-        # Set value
-        server_config[CONFIG_TEST_ARG] = True
-
-        # Test get again
-        self.assertTrue(
-            server_config[CONFIG_TEST_ARG], msg=f"{CONFIG_TEST_ARG} was not set"
-        )
-
-        # Try to set an unsupported config argument, expect failure
-        with self.assertRaises(
-            TritonModelAnalyzerException,
-            msg="Expected exception on trying to set"
-            "unsupported argument in Triton server"
-            "config",
-        ):
-            server_config["dummy"] = 1
-
-        # Reset test arg
-        server_config[CONFIG_TEST_ARG] = None
-
-        # Finally set a couple of args and then check the cli string
-        for arg, value in CLI_TO_STRING_TEST_ARGS.items():
-            server_config[arg] = value
-
-        cli_string = server_config.to_cli_string()
-        for argstring in cli_string.split():
-            # Parse the created string
-            arg, value = argstring.split("=")
-            arg = arg[2:]
-
-            # Make sure each parsed arg was in test dict
-            self.assertIn(
-                arg,
-                CLI_TO_STRING_TEST_ARGS,
-                msg=f"CLI string contained unknown argument: {arg}",
-            )
-
-            # Make sure parsed value is the one from dict, check type too
-            test_value = CLI_TO_STRING_TEST_ARGS[arg]
-            self.assertEqual(
-                test_value,
-                type(test_value)(value),
-                msg=f"CLI string contained unknown value: {value}",
-            )
-
-    def _test_create_server(self, gpus):
-        # Create a TritonServerConfig
-        server_config = TritonServerConfig()
-        server_config["model-repository"] = MODEL_REPOSITORY_PATH
-
-        # Run for both types of environments
-        self.server = TritonServerFactory.create_server_docker(
-            image=TRITON_IMAGE, config=server_config, gpus=gpus
-        )
-
-        self.server = TritonServerFactory.create_server_local(
-            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
-        )
-
-        # Try to create a server without specifying model repository and expect
-        # error
-        server_config["model-repository"] = None
-        with self.assertRaises(
-            AssertionError,
-            msg="Expected AssertionError for trying to create"
-            "server without specifying model repository.",
-        ):
-            self.server = TritonServerFactory.create_server_docker(
-                image=TRITON_IMAGE, config=server_config, gpus=gpus
-            )
-        with self.assertRaises(
-            AssertionError,
-            msg="Expected AssertionError for trying to create"
-            "server without specifying model repository.",
-        ):
-            self.server = TritonServerFactory.create_server_local(
-                path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
-            )
-
-    def test_triton_server_ssl_options(self):
-        server_config = TritonServerConfig()
-
-        triton_server_flags = {
-            "grpc-use-ssl": "1",
-            "grpc-use-ssl-mutual": "1",
-            "grpc-server-cert": "a",
-            "grpc-server-key": "b",
-            "grpc-root-cert": "c",
-        }
-        server_config.update_config(triton_server_flags)
-
-        expected_cli_str = (
-            f"--grpc-use-ssl=1 --grpc-use-ssl-mutual=1 "
-            f"--grpc-server-cert=a --grpc-server-key=b --grpc-root-cert=c"
-        )
-        self.assertEqual(server_config.to_cli_string(), expected_cli_str)
-
-    def test_create_server_no_gpu(self):
-        self._test_create_server(gpus=[])
-
-    def test_create_server_select_gpu(self):
-        self._test_create_server(gpus=self._sys_gpus[:1])
-
-    def _test_start_stop_gpus(self, gpus):
-        device_requests = [device.device_id() for device in gpus]
-
-        # Create a TritonServerConfig
-        server_config = TritonServerConfig()
-        server_config["model-repository"] = MODEL_REPOSITORY_PATH
-
-        # Create server in docker, start , wait, and stop
-        self.server = TritonServerFactory.create_server_docker(
-            image=TRITON_IMAGE, config=server_config, gpus=gpus
-        )
-
-        # Start server check that mocked api is called
-        self.server.start()
-        self.server_docker_mock.assert_server_process_start_called_with(
-            f"{TRITON_DOCKER_BIN_PATH} {server_config.to_cli_string()}",
-            MODEL_REPOSITORY_PATH,
-            TRITON_IMAGE,
-            device_requests,
-            gpus,
-            8000,
-            8001,
-            8002,
-        )
-
-        self.server_docker_mock.raise_exception_on_container_run()
-        with self.assertRaises(TritonModelAnalyzerException):
-            self.server.start()
-        self.server_docker_mock.stop_raise_exception_on_container_run()
-
-        # Stop container and check api calls
-        self.server.stop()
-        self.server_docker_mock.assert_server_process_terminate_called()
-
-        # Create local server which runs triton as a subprocess
-        self.server = TritonServerFactory.create_server_local(
-            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
-        )
-
-        # Check that API functions are called
-        self.server.start()
-
-        self.server_local_mock.assert_server_process_start_called_with(
-            cmd=[TRITON_LOCAL_BIN_PATH, "--model-repository", MODEL_REPOSITORY_PATH],
-            gpus=gpus,
-            stdout=self.server._log_file,
-        )
-
-        self.server.stop()
-        self.server_local_mock.assert_server_process_terminate_called()
-
-    def test_start_stop_gpus_no_gpu(self):
-        self._test_start_stop_gpus(gpus=[])
-
-    def test_start_stop_gpus_select_gpu(self):
-        self._test_start_stop_gpus(gpus=self._sys_gpus[:1])
-
-    def start_stop_docker_args(self):
-        device_requests, gpu_uuids = self._find_correct_gpu_settings(self._sys_gpus)
-
-        # Create a TritonServerConfig
-        server_config = TritonServerConfig()
-        server_config["model-repository"] = MODEL_REPOSITORY_PATH
-
-        # Create mounts and labels
-        mounts = ["/host/path:/dest/path:ro", "/another/host/path:/some/dest/path:rw"]
-        labels = {"RUNNER_ID": "TEST_RUNNER_ID"}
-
-        environment = {"VARIABLE": "VALUE"}
-        # Create server in docker, start , wait, and stop
-        self.server = TritonServerFactory.create_server_docker(
-            image=TRITON_IMAGE,
-            config=server_config,
-            gpus=self._sys_gpus,
-            mounts=mounts,
-            labels=labels,
-        )
-
-        # Start server check that mocked api is called
-        self.server.start(env=environment)
-        self.server_docker_mock.assert_server_process_start_called_with(
-            f"{TRITON_DOCKER_BIN_PATH} {server_config.to_cli_string()}",
-            MODEL_REPOSITORY_PATH,
-            TRITON_IMAGE,
-            device_requests,
-            gpu_uuids,
-            8000,
-            8001,
-            8002,
-            mounts,
-            labels,
-        )
-
-        # Stop container and check api calls
-        self.server.stop()
-        self.server_docker_mock.assert_server_process_terminate_called()
-
-    def _test_get_logs(self, gpus):
-        # Create a TritonServerConfig
-        server_config = TritonServerConfig()
-        server_config["model-repository"] = MODEL_REPOSITORY_PATH
-
-        # Check docker server logs
-        self.server = TritonServerFactory.create_server_docker(
-            image=TRITON_IMAGE, config=server_config, gpus=gpus
-        )
-        self.server.start()
-        self.server.stop()
-        self.server_docker_mock.assert_server_process_terminate_called()
-        self.assertEqual(self.server.logs(), "Triton Server Test Log")
-
-        # Create local server logs
-        self.server = TritonServerFactory.create_server_local(
-            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
-        )
-        self.server.start()
-        self.server.stop()
-        self.server_local_mock.assert_server_process_terminate_called()
-        self.assertEqual(self.server.logs(), "Triton Server Test Log")
-
-    def _test_cpu_stats(self, gpus):
-        device_requests = [device.device_id() for device in gpus]
-
-        # Create a TritonServerConfig
-        server_config = TritonServerConfig()
-        server_config["model-repository"] = MODEL_REPOSITORY_PATH
-
-        # Test local server cpu_stats
-        self.server = TritonServerFactory.create_server_local(
-            path=TRITON_LOCAL_BIN_PATH, config=server_config, gpus=gpus
-        )
-        self.server.start()
-        _, _ = self.server.cpu_stats()
-        self.server_local_mock.assert_cpu_stats_called()
-        self.server.stop()
-
-        # Test docker server cpu stats
-        self.server = TritonServerFactory.create_server_docker(
-            image=TRITON_IMAGE, config=server_config, gpus=gpus
-        )
-        self.server.start()
-
-        # The following needs to be called as it resets exec_run return value
-        self.server_docker_mock.assert_server_process_start_called_with(
-            f"{TRITON_DOCKER_BIN_PATH} {server_config.to_cli_string()}",
-            MODEL_REPOSITORY_PATH,
-            TRITON_IMAGE,
-            device_requests,
-            gpus,
-            8000,
-            8001,
-            8002,
-        )
-        _, _ = self.server.cpu_stats()
-        self.server_docker_mock.assert_cpu_stats_called()
-        self.server.stop()
-
-    def test_cpu_stats_no_gpu(self):
-        self._test_cpu_stats(gpus=[])
-
-    def test_cpu_stats_select_gpu(self):
-        self._test_cpu_stats(gpus=self._sys_gpus[:1])
-
-    def tearDown(self):
-        # In case test raises exception
-        if self.server is not None:
-            self.server.stop()
-
-        # Stop mocking
-        self.gpu_device_factory_mock.stop()
-        self.server_docker_mock.stop()
-        self.server_local_mock.stop()
-        self.os_mock.stop()
-        patch.stopall()
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/src/c++/perf_analyzer/tests/test_triton_server_factory.py b/src/c++/perf_analyzer/tests/test_triton_server_factory.py
deleted file mode 100755
index 2448e89c9..000000000
--- a/src/c++/perf_analyzer/tests/test_triton_server_factory.py
+++ /dev/null
@@ -1,126 +0,0 @@
-#!/usr/bin/env python3
-
-# Copyright 2022-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-from unittest.mock import MagicMock, patch
-
-from model_analyzer.config.input.config_command_profile import ConfigCommandProfile
-from model_analyzer.triton.server.server_factory import TritonServerFactory
-
-from .common import test_result_collector as trc
-from .mocks.mock_os import MockOSMethods
-
-
-class TestTritonServerFactory(trc.TestResultCollector):
-    def setUp(self):
-        # Mock path validation
-        self.mock_os = MockOSMethods(
-            mock_paths=[
-                "model_analyzer.triton.server.server_factory",
-                "model_analyzer.config.input.config_utils",
-            ]
-        )
-        self.mock_os.start()
-
-    def _test_get_server_handle_helper(
-        self, launch_mode, expect_local, expect_docker, expect_config
-    ):
-        """
-        Test get_server_handle() calls the correct Triton Server create function with the
-        correct Triton Server Config options
-
-        Parameters:
-        ----------
-        launch_mode: str
-            which launch mode to use
-        expect_local: bool
-            If true, assert that create_server_local() is called
-        expect_docker: bool
-            If true, assert that create_server_docker() is called
-        expect_config: bool
-            If true, verify that certain Triton Server args are passed to the create function as expected
-            If false, verify that those args are NOT passed
-        """
-
-        config = ConfigCommandProfile()
-        config.model_repository = "/fake_model_repository"
-        config.triton_launch_mode = launch_mode
-        config.triton_http_endpoint = "fake_address:2345"
-        config.triton_grpc_endpoint = "fake_address:4567"
-        config.monitoring_interval = 0.5
-
-        expected_http_port = "2345"
-        expected_grpc_port = "4567"
-        # Convert seconds to ms
-        expected_metrics_interval_ms = int(config.monitoring_interval * 1000)
-
-        with patch(
-            "model_analyzer.triton.server.server_factory.TritonServerFactory.create_server_local"
-        ) as mocked_local, patch(
-            "model_analyzer.triton.server.server_factory.TritonServerFactory.create_server_docker"
-        ) as mocked_docker, patch(
-            "model_analyzer.triton.server.server_factory.TritonServerFactory._validate_triton_install_path"
-        ), patch(
-            "model_analyzer.triton.server.server_factory.TritonServerFactory._validate_triton_server_path"
-        ):
-            _ = TritonServerFactory.get_server_handle(config, MagicMock(), False)
-            self.assertEqual(mocked_local.call_count, expect_local)
-            self.assertEqual(mocked_docker.call_count, expect_docker)
-
-            if expect_local:
-                _, kwargs = mocked_local.call_args
-            else:
-                _, kwargs = mocked_docker.call_args
-            triton_config = kwargs["config"]
-
-            if expect_config:
-                self.assertEqual(triton_config["http-port"], expected_http_port)
-                self.assertEqual(triton_config["grpc-port"], expected_grpc_port)
-                self.assertEqual(
-                    triton_config["metrics-interval-ms"], expected_metrics_interval_ms
-                )
-
-            else:
-                self.assertEqual(triton_config["http-port"], None)
-                self.assertEqual(triton_config["grpc-port"], None)
-                self.assertEqual(triton_config["metrics-interval-ms"], None)
-
-    def test_get_server_handle_remote(self):
-        self._test_get_server_handle_helper(
-            launch_mode="remote", expect_local=1, expect_docker=0, expect_config=False
-        )
-
-    def test_get_server_handle_c_api(self):
-        self._test_get_server_handle_helper(
-            launch_mode="c_api", expect_local=1, expect_docker=0, expect_config=False
-        )
-
-    def test_get_server_handle_local(self):
-        self._test_get_server_handle_helper(
-            launch_mode="local", expect_local=1, expect_docker=0, expect_config=True
-        )
-
-    def test_get_server_handle_docker(self):
-        self._test_get_server_handle_helper(
-            launch_mode="docker", expect_local=0, expect_docker=1, expect_config=True
-        )
-
-    def tearDown(self):
-        patch.stopall()
-
-
-if __name__ == "__main__":
-    unittest.main()

From f382bffa0548bf5ab2f7afd26047de3bea0ca610 Mon Sep 17 00:00:00 2001
From: David Yastremsky <dyastremsky@nvidia.com>
Date: Mon, 26 Feb 2024 15:59:48 -0800
Subject: [PATCH 3/4] Run pre-commit

---
 src/c++/perf_analyzer/tests/test_cli.py          | 1 -
 src/c++/perf_analyzer/tests/test_yaml_options.py | 1 -
 2 files changed, 2 deletions(-)

diff --git a/src/c++/perf_analyzer/tests/test_cli.py b/src/c++/perf_analyzer/tests/test_cli.py
index 98ec60237..55400e1d0 100755
--- a/src/c++/perf_analyzer/tests/test_cli.py
+++ b/src/c++/perf_analyzer/tests/test_cli.py
@@ -29,7 +29,6 @@
 from unittest.mock import MagicMock, patch
 
 import psutil
-
 from model_analyzer.cli.cli import CLI
 from model_analyzer.config.input.config_command_profile import ConfigCommandProfile
 from model_analyzer.config.input.config_command_report import ConfigCommandReport
diff --git a/src/c++/perf_analyzer/tests/test_yaml_options.py b/src/c++/perf_analyzer/tests/test_yaml_options.py
index 1c7ce9a4d..52f2e2da8 100755
--- a/src/c++/perf_analyzer/tests/test_yaml_options.py
+++ b/src/c++/perf_analyzer/tests/test_yaml_options.py
@@ -15,7 +15,6 @@
 # limitations under the License.
 
 import yaml
-
 from model_analyzer.config.input.yaml_config_validator import YamlConfigValidator
 from model_analyzer.model_analyzer_exceptions import TritonModelAnalyzerException
 

From 15a5beef638efe7af9585f0b0b62837cbaee0b61 Mon Sep 17 00:00:00 2001
From: Ryan McCormick <rmccormick@nvidia.com>
Date: Tue, 27 Feb 2024 14:54:14 -0800
Subject: [PATCH 4/4] Add boilerplate for parsing, testing, and wrapping PA
 (#471)

* Add boilerplate code with placeholder for running PA

* Add return value for codeql
---
 src/c++/perf_analyzer/genai-pa/README.md      |  30 ++++
 .../genai-pa/genai_pa/.gitignore              |   2 +
 .../genai-pa/genai_pa/__init__.py             |  27 ++++
 .../genai-pa/genai_pa/constants.py            |  28 ++++
 .../perf_analyzer/genai-pa/genai_pa/main.py   |  58 +++++++
 .../perf_analyzer/genai-pa/genai_pa/parser.py | 146 ++++++++++++++++++
 .../genai-pa/genai_pa/wrapper.py              |  56 +++++++
 src/c++/perf_analyzer/genai-pa/pyproject.toml |  85 ++++++++++
 .../perf_analyzer/genai-pa/tests/test_cli.py  |  37 +++++
 .../genai-pa/tests/test_library.py            |  32 ++++
 10 files changed, 501 insertions(+)
 create mode 100644 src/c++/perf_analyzer/genai-pa/README.md
 create mode 100644 src/c++/perf_analyzer/genai-pa/genai_pa/.gitignore
 create mode 100644 src/c++/perf_analyzer/genai-pa/genai_pa/__init__.py
 create mode 100644 src/c++/perf_analyzer/genai-pa/genai_pa/constants.py
 create mode 100755 src/c++/perf_analyzer/genai-pa/genai_pa/main.py
 create mode 100755 src/c++/perf_analyzer/genai-pa/genai_pa/parser.py
 create mode 100644 src/c++/perf_analyzer/genai-pa/genai_pa/wrapper.py
 create mode 100644 src/c++/perf_analyzer/genai-pa/pyproject.toml
 create mode 100644 src/c++/perf_analyzer/genai-pa/tests/test_cli.py
 create mode 100644 src/c++/perf_analyzer/genai-pa/tests/test_library.py

diff --git a/src/c++/perf_analyzer/genai-pa/README.md b/src/c++/perf_analyzer/genai-pa/README.md
new file mode 100644
index 000000000..07f17bf2d
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/README.md
@@ -0,0 +1,30 @@
+# genai-pa
+
+## Installation
+
+### Install from Source
+
+```bash
+pip install .
+```
+
+## Quickstart
+
+```bash
+# Explore the commands
+genai-pa -h
+```
+
+## Examples
+
+```
+# Profile an LLM with hard-coded inputs
+genai-pa -m my_llm_model
+```
+
+## Test
+
+```
+pip install .
+pytest tests/
+```
diff --git a/src/c++/perf_analyzer/genai-pa/genai_pa/.gitignore b/src/c++/perf_analyzer/genai-pa/genai_pa/.gitignore
new file mode 100644
index 000000000..973a71df2
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/genai_pa/.gitignore
@@ -0,0 +1,2 @@
+*.json
+*.cache
diff --git a/src/c++/perf_analyzer/genai-pa/genai_pa/__init__.py b/src/c++/perf_analyzer/genai-pa/genai_pa/__init__.py
new file mode 100644
index 000000000..4d50bae64
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/genai_pa/__init__.py
@@ -0,0 +1,27 @@
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+__version__ = "0.0.1"
diff --git a/src/c++/perf_analyzer/genai-pa/genai_pa/constants.py b/src/c++/perf_analyzer/genai-pa/genai_pa/constants.py
new file mode 100644
index 000000000..09686f38f
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/genai_pa/constants.py
@@ -0,0 +1,28 @@
+#!/usr/bin/env python3
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+LOGGER_NAME: str = "genai-pa"
diff --git a/src/c++/perf_analyzer/genai-pa/genai_pa/main.py b/src/c++/perf_analyzer/genai-pa/genai_pa/main.py
new file mode 100755
index 000000000..79748df50
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/genai_pa/main.py
@@ -0,0 +1,58 @@
+#!/usr/bin/env python3
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+import logging
+import sys
+
+from genai_pa import parser
+from genai_pa.constants import LOGGER_NAME
+
+logging.basicConfig(level=logging.INFO, format="%(name)s - %(levelname)s - %(message)s")
+logger = logging.getLogger(LOGGER_NAME)
+
+
+# Separate function that can raise exceptions used for testing
+# to assert correct errors and messages.
+# Optional argv used for testing - will default to sys.argv if None.
+def run(argv=None):
+    args = parser.parse_args(argv)
+    args.func(args)
+
+
+def main():
+    # Interactive use will catch exceptions and log formatted errors rather than tracebacks.
+    try:
+        run()
+    except Exception as e:
+        logger.error(f"{e}")
+        return 1
+
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
diff --git a/src/c++/perf_analyzer/genai-pa/genai_pa/parser.py b/src/c++/perf_analyzer/genai-pa/genai_pa/parser.py
new file mode 100755
index 000000000..236891b78
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/genai_pa/parser.py
@@ -0,0 +1,146 @@
+#!/usr/bin/env python3
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+import argparse
+import logging
+
+from genai_pa.constants import LOGGER_NAME
+
+logger = logging.getLogger(LOGGER_NAME)
+
+### Handlers ###
+
+
+# NOTE: Placeholder
+def handler(args):
+    from genai_pa.wrapper import Profiler
+
+    Profiler.run(
+        model=args.model,
+    )
+
+
+### Parsers ###
+
+
+def add_model_args(parser):
+    parser.add_argument(
+        "-m",
+        "--model",
+        type=str,
+        required=True,
+        help=f"The name of the model to benchmark.",
+    )
+
+
+def add_profile_args(parser):
+    parser.add_argument(
+        "-b",
+        "--batch-size",
+        type=int,
+        default=1,
+        required=False,
+        help="The batch size / concurrency to benchmark. (Default: 1)",
+    )
+    parser.add_argument(
+        "--input-length",
+        type=int,
+        default=128,
+        required=False,
+        help="The input length (tokens) to use for benchmarking LLMs. (Default: 128)",
+    )
+    parser.add_argument(
+        "--output-length",
+        type=int,
+        default=128,
+        required=False,
+        help="The output length (tokens) to use for benchmarking LLMs. (Default: 128)",
+    )
+
+
+def add_endpoint_args(parser):
+    parser.add_argument(
+        "--url",
+        type=str,
+        default="localhost:8001",
+        required=False,
+        help="URL of the endpoint to target for benchmarking.",
+    )
+    parser.add_argument(
+        "--provider",
+        type=str,
+        choices=["triton", "openai"],
+        required=False,
+        help="Provider format/schema to use for benchmarking.",
+    )
+
+
+def add_dataset_args(parser):
+    parser.add_argument(
+        "--dataset",
+        type=str,
+        default="OpenOrca",
+        choices=["OpenOrca", "cnn_dailymail"],
+        required=False,
+        help="HuggingFace dataset to use for the benchmark.",
+    )
+    parser.add_argument(
+        "--tokenizer",
+        type=str,
+        default="auto",
+        choices=["auto"],
+        required=False,
+        help="The HuggingFace tokenizer to use to interpret token metrics from final text results",
+    )
+
+
+### Entrypoint ###
+
+
+# Optional argv used for testing - will default to sys.argv if None.
+def parse_args(argv=None):
+    parser = argparse.ArgumentParser(
+        prog="genai-pa",
+        description="CLI to profile LLMs and Generative AI models with PA",
+    )
+    parser.set_defaults(func=handler)
+
+    # Conceptually group args for easier visualization
+    model_group = parser.add_argument_group("Model")
+    add_model_args(model_group)
+
+    profile_group = parser.add_argument_group("Profiling")
+    add_profile_args(profile_group)
+
+    endpoint_group = parser.add_argument_group("Endpoint")
+    add_endpoint_args(endpoint_group)
+
+    dataset_group = parser.add_argument_group("Dataset")
+    add_dataset_args(dataset_group)
+
+    args = parser.parse_args(argv)
+    return args
diff --git a/src/c++/perf_analyzer/genai-pa/genai_pa/wrapper.py b/src/c++/perf_analyzer/genai-pa/genai_pa/wrapper.py
new file mode 100644
index 000000000..8e466b1fd
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/genai_pa/wrapper.py
@@ -0,0 +1,56 @@
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+import json
+import logging
+import subprocess
+
+from genai_pa.constants import LOGGER_NAME
+
+logger = logging.getLogger(LOGGER_NAME)
+
+
+class Profiler:
+    @staticmethod
+    def run(model):
+        # TODO: Replace with other plumbing
+        input_file = "/tmp/input_data.json"
+        with open(input_file, "w") as f:
+            data = {"data": [{"text_input": ["hi"]}]}
+            json.dump(data, f)
+
+        cmd = [
+            "perf_analyzer",
+            "-i",
+            "grpc",
+            "--streaming",
+            "-m",
+            model,
+            "--input-data",
+            input_file,
+        ]
+        logger.info(f"Running Perf Analyzer : '{cmd}'")
+        subprocess.run(cmd)
diff --git a/src/c++/perf_analyzer/genai-pa/pyproject.toml b/src/c++/perf_analyzer/genai-pa/pyproject.toml
new file mode 100644
index 000000000..f6bfc850c
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/pyproject.toml
@@ -0,0 +1,85 @@
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+[project]
+name = "genai-pa"
+readme = "README.md"
+description = "GenAI Perf Analyzer CLI - CLI tool to simplify profiling LLMs and Generative AI models with Perf Analyzer"
+dynamic = ["version"]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Science/Research",
+    "Intended Audience :: Developers",
+    "Topic :: Software Development",
+    "Topic :: Scientific/Engineering",
+    "Programming Language :: Python",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.8",
+    "Programming Language :: Python :: 3.10",
+    "Operating System :: Unix",
+]
+authors = []
+maintainers = []
+keywords = []
+requires-python = ">=3.8,<4"
+dependencies = [
+  "numpy",
+  "rich"
+]
+
+# CLI Entrypoint
+[project.scripts]
+genai-pa = "genai_pa.main:main"
+
+[project.urls]
+"Homepage" = "https://github.com/triton-inference-server/"
+"Bug Tracker" = "https://github.com/triton-inference-server/server/issues"
+
+# Build
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.version]
+path = "genai_pa/__init__.py"
+
+# Pre-commit hook tool configs
+[tool.codespell]
+# note: pre-commit passes explicit lists of files here, which this skip file list doesn't override -
+# this is only to allow you to run codespell interactively
+skip = "./.git,./.github"
+# ignore short words, and typename parameters like OffsetT
+ignore-regex = "\\b(.{1,4}|[A-Z]\\w*T)\\b"
+# use the 'clear' dictionary for unambiguous spelling mistakes
+builtin = "clear"
+# disable warnings about binary files and wrong encoding
+quiet-level = 3
+
+# Linting/formatting
+[tool.ruff]
+# Same as Black.
+line-length = 88
+indent-width = 4
diff --git a/src/c++/perf_analyzer/genai-pa/tests/test_cli.py b/src/c++/perf_analyzer/genai-pa/tests/test_cli.py
new file mode 100644
index 000000000..cddbb70f9
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/tests/test_cli.py
@@ -0,0 +1,37 @@
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+import pytest
+from genai_pa.main import run
+
+
+# NOTE: Placeholder
+class TestHelp:
+    @pytest.mark.parametrize("arg", ["-h", "--help"])
+    def test_help(self, arg):
+        args = [arg]
+        with pytest.raises(SystemExit):
+            run(args)
diff --git a/src/c++/perf_analyzer/genai-pa/tests/test_library.py b/src/c++/perf_analyzer/genai-pa/tests/test_library.py
new file mode 100644
index 000000000..36c7624b1
--- /dev/null
+++ b/src/c++/perf_analyzer/genai-pa/tests/test_library.py
@@ -0,0 +1,32 @@
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+import genai_pa
+
+
+# Placeholder to add real tests in the future
+def test_version():
+    print(genai_pa.__version__)