Build Instructions¶

Note: The most up-to-date build instructions are embedded in a set of scripts bundled in the FBGEMM repo under setup_env.bash.

The currently available FBGEMM GenAI build variants are:

CUDA

The general steps for building FBGEMM GenAI are as follows:

Set up an isolated build environment.
Set up the toolchain for either a CUDA build.
Install PyTorch.
Run the build script.

Set Up an Isolated Build Environment¶

Follow the instructions to set up the Conda environment:

Install CUTLASS¶

CUTLASS should be already be available in the FBGEMM repository as a git submodule (see Preparing the Build). The following include paths are already added to the CMake configuration:

set(THIRDPARTY ${FBGEMM}/external)

${THIRDPARTY}/cutlass/include
${THIRDPARTY}/cutlass/tools/util/include

Other Pre-Build Setup¶

As FBGEMM GenAI leverages the same build process as FBGEMM_GPU, please refer to Preparing the Build for additional pre-build setup information.

Preparing the Build¶

Clone the repo along with its submodules, and install requirements_genai.txt:

# !! Run inside the Conda environment !!

# Select a version tag
FBGEMM_VERSION=v1.2.0

# Clone the repo along with its submodules
git clone --recursive -b ${FBGEMM_VERSION} https://github.com/pytorch/FBGEMM.git fbgemm_${FBGEMM_VERSION}

# Install additional required packages for building and testing
cd fbgemm_${FBGEMM_VERSION}/fbgemm_gpu
pip install -r requirements_genai.txt

Set Wheel Build Variables¶

When building out the Python wheel, the package name, Python version tag, and Python platform name must first be properly set:

# Set the package name depending on the build variant
export package_name=fbgemm_genai_{cuda}

# Set the Python version tag.  It should follow the convention `py<major><minor>`,
# e.g. Python 3.13 --> py313
export python_tag=py313

# Determine the processor architecture
export ARCH=$(uname -m)

# Set the Python platform name for the Linux case
export python_plat_name="manylinux_2_28_${ARCH}"
# For the macOS (x86_64) case
export python_plat_name="macosx_10_9_${ARCH}"
# For the macOS (arm64) case
export python_plat_name="macosx_11_0_${ARCH}"
# For the Windows case
export python_plat_name="win_${ARCH}"

CUDA Build¶

Building FBGEMM GenAI for CUDA requires both NVML and cuDNN to be installed and made available to the build through environment variables. The presence of a CUDA device, however, is not required for building the package.

Similar to CPU-only builds, building with Clang + libstdc++ can be enabled by appending --cxxprefix=$CONDA_PREFIX to the build command, presuming the toolchains have been properly installed.

# !! Run in fbgemm_gpu/ directory inside the Conda environment !!

# [OPTIONAL] Specify the CUDA installation paths
# This may be required if CMake is unable to find nvcc
export CUDACXX=/path/to/nvcc
export CUDA_BIN_PATH=/path/to/cuda/installation

# [OPTIONAL] Provide the CUB installation directory (applicable only to CUDA versions prior to 11.1)
export CUB_DIR=/path/to/cub

# [OPTIONAL] Allow NVCC to use host compilers that are newer than what NVCC officially supports
nvcc_prepend_flags=(
  -allow-unsupported-compiler
)

# [OPTIONAL] If clang is the host compiler, set NVCC to use libstdc++ since libc++ is not supported
nvcc_prepend_flags+=(
  -Xcompiler -stdlib=libstdc++
  -ccbin "/path/to/clang++"
)

# [OPTIONAL] Set NVCC_PREPEND_FLAGS as needed
export NVCC_PREPEND_FLAGS="${nvcc_prepend_flags[@]}"

# [OPTIONAL] Enable verbose NVCC logs
export NVCC_VERBOSE=1

# Specify cuDNN header and library paths
export CUDNN_INCLUDE_DIR=/path/to/cudnn/include
export CUDNN_LIBRARY=/path/to/cudnn/lib

# Specify NVML filepath
export NVML_LIB_PATH=/path/to/libnvidia-ml.so

# Specify NCCL filepath
export NCCL_LIB_PATH=/path/to/libnccl.so.2

# Build for SM70/80 (V100/A100 GPU); update as needed
# If not specified, only the CUDA architecture supported by current system will be targeted
# If not specified and no CUDA device is present either, all CUDA architectures will be targeted
cuda_arch_list=7.0;8.0

# Unset TORCH_CUDA_ARCH_LIST if it exists, bc it takes precedence over
# -DTORCH_CUDA_ARCH_LIST during the invocation of setup.py
unset TORCH_CUDA_ARCH_LIST

# Build the wheel artifact only
python setup.py bdist_wheel \
    --package_variant=genai \
    --python-tag="${python_tag}" \
    --plat-name="${python_plat_name}" \
    --nvml_lib_path=${NVML_LIB_PATH} \
    --nccl_lib_path=${NCCL_LIB_PATH} \
    -DTORCH_CUDA_ARCH_LIST="${cuda_arch_list}"

# Build and install the library into the Conda environment
python setup.py install \
    --package_variant=genai \
    --nvml_lib_path=${NVML_LIB_PATH} \
    --nccl_lib_path=${NCCL_LIB_PATH} \
    -DTORCH_CUDA_ARCH_LIST="${cuda_arch_list}"

Post-Build Checks (For Developers)¶

As FBGEMM GenAI leverages the same build process as FBGEMM_GPU, please refer to Post-Build Checks (For Developers) for information on additional post-build checks.