Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add boilerplate for parsing, testing, and wrapping PA #471

Merged
merged 2 commits into from
Feb 27, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 30 additions & 0 deletions src/c++/perf_analyzer/genai-pa/README.md
Original file line number Diff line number Diff line change
@@ -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/
```
2 changes: 2 additions & 0 deletions src/c++/perf_analyzer/genai-pa/genai_pa/.gitignore
Original file line number Diff line number Diff line change
@@ -0,0 +1,2 @@
*.json
*.cache
27 changes: 27 additions & 0 deletions src/c++/perf_analyzer/genai-pa/genai_pa/__init__.py
Original file line number Diff line number Diff line change
@@ -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"
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,4 @@
# (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 sys

args = parser.parse_args()
if __name__ == "__main__":
sys.exit(0)
LOGGER_NAME: str = "genai-pa"
58 changes: 58 additions & 0 deletions src/c++/perf_analyzer/genai-pa/genai_pa/main.py
Original file line number Diff line number Diff line change
@@ -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())
Fixed Show fixed Hide fixed
146 changes: 146 additions & 0 deletions src/c++/perf_analyzer/genai-pa/genai_pa/parser.py
Original file line number Diff line number Diff line change
@@ -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
56 changes: 56 additions & 0 deletions src/c++/perf_analyzer/genai-pa/genai_pa/wrapper.py
Original file line number Diff line number Diff line change
@@ -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)
Loading
Loading