[LoadGen] Simplify installation with a pip one-liner

loadgen/README_BUILD.md

@@ -3,18 +3,22 @@
 ## Prerequisites
 
     sudo apt-get install libglib2.0-dev python-pip python3-pip
-    pip2 install absl-py numpy
-    pip3 install absl-py numpy
 
 ## Quick Start
 ### Installation - Python
+If you need to clone the repo (e.g., because you are a MLPerf Inference developer), you
+can build and install the `mlperf-loadgen` package via:
 
-    pip install absl-py numpy
     git clone --recurse-submodules https://github.com/mlcommons/inference.git mlperf_inference
     cd mlperf_inference/loadgen
-    CFLAGS="-std=c++14 -O3" python -m pip install .
+    pip install -e .  # Install in the editable mode because you are a developer.
 
-This will fetch the loadgen source, build and install the loadgen as a python module, and run a simple end-to-end demo.
+If you don't need to clone the repo (e.g., you just want to install `mlperf-loadgen`
+from the latest commit of the `master` branch):
+
+    pip install git+https://github.com/mlcommons/inference.git#subdirectory=loadgen
+
+This will fetch the loadgen source, then build and install the loadgen as a python module.
 
 Alternatively, we provide wheels for several python versions and operating system that can be installed using pip directly.
 

loadgen/requirements.txt

@@ -1 +1,3 @@
 pybind11
+absl-py
+numpy

loadgen/setup.py

@@ -24,12 +24,12 @@
 #  and binaries. Use one of the gn build targets instead if you want
 #  to avoid poluting the source tree.
 
-from setuptools import Extension, setup
 from pathlib import Path
+
 from pybind11 import get_include
 from pybind11.setup_helpers import Pybind11Extension, build_ext
+from setuptools import setup
 from version_generator import generate_loadgen_version_definitions
-import subprocess
 
 generated_version_source_filename = "generated/version_generated.cc"
 generate_loadgen_version_definitions(generated_version_source_filename, ".")
@@ -42,7 +42,7 @@
     "test_settings.h",
     "issue_query_controller.h",
     "early_stopping.h",
-    "query_dispatch_library.h"
+    "query_dispatch_library.h",
 ]
 
 lib_headers = [
@@ -54,7 +54,7 @@
     "results.h",
     "bindings/c_api.h",
     "version_generator.py",
-    "mlperf_conf.h"
+    "mlperf_conf.h",
 ]
 
 lib_sources = [
@@ -91,20 +91,29 @@
 if len(version_split) < 2:
     print("Version is incomplete. Needs a format like 4.1.1 in VERSION file")
 
+# Read requirements from requirements.txt
+install_requires = []
+requirements_file = this_directory / "requirements.txt"
+if requirements_file.exists():
+    with open(requirements_file, "r") as f:
+        install_requires = [
+            line.strip() for line in f if line.strip() and not line.startswith("#")
+        ]
+
 
 try:
-    with open("mlperf.conf", 'r') as file:
+    with open("mlperf.conf", "r") as file:
         conf_contents = file.read()
 
     # Escape backslashes and double quotes
-    conf_contents = conf_contents.replace('\\', '\\\\').replace('"', '\\"')
+    conf_contents = conf_contents.replace("\\", "\\\\").replace('"', '\\"')
 
     # Convert newlines
-    conf_contents = conf_contents.replace('\n', '\\n"\n"')
+    conf_contents = conf_contents.replace("\n", '\\n"\n"')
 
     formatted_content = f'const char* mlperf_conf =\n"{conf_contents}";\n'
 
-    with open("mlperf_conf.h", 'w') as header_file:
+    with open("mlperf_conf.h", "w") as header_file:
         header_file.write(formatted_content)
 
 except IOError as e:
@@ -113,24 +122,26 @@
 mlperf_loadgen_module = Pybind11Extension(
     "mlperf_loadgen",
     define_macros=[
-        ("MAJOR_VERSION",
-         version_split[0]),
-        ("MINOR_VERSION",
-         version_split[1])
+        ("MAJOR_VERSION", version_split[0]),
+        ("MINOR_VERSION", version_split[1]),
     ],
     include_dirs=[".", get_include()],
     sources=mlperf_loadgen_sources,
     depends=mlperf_loadgen_headers,
+    extra_compile_args=["-std=c++14", "-O3"],
 )
 
-setup(name="mlcommons_loadgen",
-      version=version,
-      description="MLPerf Inference LoadGen python bindings",
-      url="https://mlcommons.org/",
-      cmdclass={"build_ext": build_ext},
-      ext_modules=[mlperf_loadgen_module],
-      packages=['mlcommons_loadgen'],
-      package_dir={'mlcommons_loadgen': '.'},
-      include_package_data=True,
-      long_description=mlperf_long_description,
-      long_description_content_type='text/markdown')
+setup(
+    name="mlcommons_loadgen",
+    version=version,
+    description="MLPerf Inference LoadGen python bindings",
+    url="https://mlcommons.org/",
+    cmdclass={"build_ext": build_ext},
+    ext_modules=[mlperf_loadgen_module],
+    packages=["mlcommons_loadgen"],
+    package_dir={"mlcommons_loadgen": "."},
+    include_package_data=True,
+    install_requires=install_requires,
+    long_description=mlperf_long_description,
+    long_description_content_type="text/markdown",
+)

multimodal/vl2l/README.md

@@ -2,20 +2,6 @@
 
 ## Quick Start
 
-### Get the source code 
-
-Clone the MLPerf Inference repo via:
-
-```bash
-git clone --recurse-submodules https://github.com/mlcommons/inference.git mlperf-inference
-```
-
-Then enter the repo: 
-
-```bash
-cd mlperf-inference/
-```
-
 ### Create a Conda environment
 
 Follow [this link](https://www.anaconda.com/docs/getting-started/miniconda/install#quickstart-install-instructions)
@@ -26,53 +12,32 @@ environment via:
 conda create -n mlperf-inf-mm-vl2l python=3.12
 ```
 
-### Install LoadGen
-
-Update `libstdc++` in the conda environment:
+### (Optionally) Update `libstdc++` in the conda environment:
 
 ```bash
 conda install -c conda-forge libstdcxx-ng
 ```
 
-Install `absl-py` and `numpy`:
-
-```bash
-conda install absl-py numpy
-```
-
-Build and install LoadGen from source:
-
-```bash
-cd loadgen/
-CFLAGS="-std=c++14 -O3" python -m pip install .
-cd ../
-```
-
-Run a quick test to validate that LoadGen was installed correctly:
-
-```bash
-python loadgen/demos/token_metrics/py_demo_server.py
-```
+This is only needed when your local environment doesn't have `libstdc++`.
 
 ### Install the VL2L benchmarking CLI
 
 For users, install `mlperf-inf-mm-vl2l` with:
 
 ```bash
-pip install multimodal/vl2l/
+pip install git+https://github.com/mlcommons/inference.git#subdirectory=multimodal/vl2l
 ```
 
 For developers, install `mlperf-inf-mm-vl2l` and the development tools with:
-
-- On Bash
+1. Clone the MLPerf Inference repo.
 ```bash
-pip install multimodal/vl2l/[dev]
-```
-- On Zsh
-```zsh
-pip install multimodal/vl2l/"[dev]"
+git clone --recurse-submodules https://github.com/mlcommons/inference.git mlperf-inference
 ```
 
+2. Install in editable mode with the development tools.
+- Bash: `pip install -e mlperf-inference/multimodal/vl2l/[dev]`
+- Zsh: `pip install -e mlperf-inference/multimodal/vl2l/"[dev]"`
+
 After installation, you can check the CLI flags that `mlperf-inf-mm-vl2l` can take with:
 
 ```bash

multimodal/vl2l/pyproject.toml

@@ -13,7 +13,7 @@ dependencies = [
   "datasets",
   "loguru",
   "matplotlib",
-  "mlcommons_loadgen",
+  "mlcommons_loadgen @ git+https://github.com/mlcommons/inference.git#subdirectory=loadgen",
   "openai[aiohttp]",
   "pydantic",
   "pydantic-typer @ git+https://github.com/CentML/pydantic-typer.git@wangshangsam/preserve-full-annotated-type",