[EP Perf] Fix missing Azure cli & use onnx zoo model inside image (#1…

…8917)

### Description
* Fix [missing Azure CLI
issue](https://aiinfra.visualstudio.com/Lotus/_build/results?buildId=392612&view=logs&j=b6bfa4e2-8141-507f-8ca1-59b3f929fa71&t=d0fed32c-7043-5439-8bf2-dd69d21beb5b&l=12).
* Now, once CI fails to run `az --version`, it would auto-reinstall the
azure cli dependency
* Use existing onnx zoo model inside image during memtesting 
   * to avoid test failure when onnx model zoo is restructuring
* Display more detail info of valgrind when memtesting
* Clear invalid dep of existing AddressSanitizer test case


### Validate
* Before the fix, Azure CLI is missing:
https://aiinfra.visualstudio.com/Lotus/_build/results?buildId=392994&view=logs&j=b6bfa4e2-8141-507f-8ca1-59b3f929fa71&t=d0fed32c-7043-5439-8bf2-dd69d21beb5b&l=10
* After the fix:
https://aiinfra.visualstudio.com/Lotus/_build/results?buildId=392619&view=logs&j=b6bfa4e2-8141-507f-8ca1-59b3f929fa71&t=d0fed32c-7043-5439-8bf2-dd69d21beb5b

onnxruntime/python/tools/tensorrt/perf/mem_test/main.cpp

@@ -6,7 +6,6 @@
 #include <iostream>
 #include <vector>
 #include <onnxruntime_cxx_api.h>
-#include <tensorrt_provider_factory.h>
 #include <tensorrt_provider_options.h>
 
 void run_ort_trt2() {
@@ -135,7 +134,7 @@ void run_ort_trt() {
 
   session_options.SetGraphOptimizationLevel(GraphOptimizationLevel::ORT_ENABLE_EXTENDED);
 
-  const char* model_path = "squeezenet.onnx";
+  const char* model_path = "/data/ep-perf-models/onnx-zoo-models/squeezenet1.0-7/squeezenet/model.onnx";
 
   Ort::ThrowOnError(api.CreateTensorRTProviderOptions(&tensorrt_options));
   std::unique_ptr<OrtTensorRTProviderOptionsV2, decltype(api.ReleaseTensorRTProviderOptions)> rel_trt_options(

onnxruntime/python/tools/tensorrt/perf/mem_test/run.sh

@@ -14,9 +14,7 @@ s) ORT_SOURCE=${OPTARG};;
 esac
 done
 
-ONNX_MODEL_TAR_URL="https://github.com/onnx/models/raw/main/archive/vision/classification/squeezenet/model/squeezenet1.0-7.tar.gz"
-MODEL_TAR_NAME="squeezenet1.0-7.tar.gz"
-ONNX_MODEL="squeezenet.onnx"
+ONNX_MODEL="/data/ep-perf-models/onnx-zoo-models/squeezenet1.0-7/squeezenet/model.onnx"
 ASAN_OPTIONS="protect_shadow_gap=0:new_delete_type_mismatch=0:log_path=asan.log"
 
 export LD_LIBRARY_PATH=${ORT_BINARY_PATH}
@@ -48,15 +46,11 @@ cp ../squeezenet_calibration.flatbuffers .
 
 cmake ..
 make -j
-wget ${ONNX_MODEL_TAR_URL} -O squeezenet1.0-7.tar.gz
-tar -xzf ${MODEL_TAR_NAME} --strip-components=1
-mv model.onnx ${ONNX_MODEL}
-rm ${MODEL_TAR_NAME}
 mkdir result
 
 # Run valgrind
 echo $(date +"%Y-%m-%d %H:%M:%S") '[valgrind] Starting memcheck with' ${ONNX_MODEL}
-valgrind --leak-check=full --show-leak-kinds=all --log-file=valgrind.log ${ORT_SOURCE}/build/Linux/Release/onnxruntime_perf_test -e tensorrt -r 1 ${ONNX_MODEL}
+valgrind --leak-check=full --show-leak-kinds=definite --max-threads=3000 --num-callers=20 --keep-debuginfo=yes --log-file=valgrind.log ${ORT_SOURCE}/build/Linux/Release/onnxruntime_perf_test -e tensorrt -r 1 ${ONNX_MODEL}
 echo $(date +"%Y-%m-%d %H:%M:%S") '[valgrind] Analyzing valgrind log'
 
 found_leak_summary=false

onnxruntime/python/tools/tensorrt/perf/mem_test/run_mem_test_docker.sh

@@ -24,4 +24,4 @@ then
     BUILD_ORT_LATEST="true"
 fi
 
-docker run --rm --gpus all -v $MEM_TEST_DIR:$DOCKER_MEM_TEST_DIR $DOCKER_IMAGE /bin/bash $DOCKER_MEM_TEST_DIR'run.sh' -p $DOCKER_MEM_TEST_DIR -o $DOCKER_ORT_LIBS -s $DOCKER_ORT_SOURCE -l $BUILD_ORT_LATEST
+docker run --rm --gpus all -v $MEM_TEST_DIR:$DOCKER_MEM_TEST_DIR -v /data/ep-perf-models:/data/ep-perf-models $DOCKER_IMAGE /bin/bash $DOCKER_MEM_TEST_DIR'run.sh' -p $DOCKER_MEM_TEST_DIR -o $DOCKER_ORT_LIBS -s $DOCKER_ORT_SOURCE -l $BUILD_ORT_LATEST

tools/ci_build/github/azure-pipelines/linux-gpu-tensorrt-daily-perf-pipeline.yml

@@ -126,6 +126,13 @@ jobs:
       - script: 'python3 -m pip install pandas azure-kusto-data[pandas] azure-kusto-ingest[pandas] coloredlogs'
         displayName: 'Install dashboard dependencies'
 
+      - script: |
+          az --version || {
+              echo "Azure CLI not found, installing..."
+              curl -sL https://aka.ms/InstallAzureCLIDeb | sudo bash
+          }
+        displayName: 'Check and Install Azure CLI'
+
       - task: AzureCLI@2
         displayName: 'Azure CLI Post to Dashboard'
         inputs: