Skip to content

Introduce sve function for matrix multiplication #752

Introduce sve function for matrix multiplication

Introduce sve function for matrix multiplication #752

# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
# This source code is licensed under the BSD-style license found in the
# LICENSE file in the root directory of this source tree.
# This workflow is used for FBGEMM_GPU-GenAI CI, and is meant to be used for
# copies of the FBGEMM repos hosted outside of the pytorch org.
name: FBGEMM_GPU-GenAI CI (Generic Runner)
on:
# PR Trigger
#
pull_request:
branches:
- main
# Push Trigger (enable to catch errors coming out of multiple merges)
#
push:
branches:
- main
# Manual Trigger
#
workflow_dispatch:
concurrency:
# Cancel previous runs in the PR if a new commit is pushed
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
jobs:
# Build on CPU hosts and upload to GHA
build_artifact:
if: ${{ github.repository_owner != 'pytorch' }}
runs-on: ${{ matrix.host-machine.instance }}
container:
image: amazonlinux:2023
options: --user root --privileged --pid=host
volumes:
- /var/run/docker.sock:/var/run/docker.sock
defaults:
run:
shell: bash
env:
PRELUDE: .github/scripts/setup_env.bash
BUILD_ENV: build_binary
BUILD_VARIANT: genai
continue-on-error: true
strategy:
fail-fast: false
matrix:
host-machine: [
{ arch: x86, instance: "ubuntu-latest" },
]
python-version: [ "3.9", "3.10", "3.11", "3.12" ]
cuda-version: [ "11.8.0", "12.1.1", "12.4.1" ]
compiler: [ "gcc", "clang" ]
steps:
- name: Setup Build Container
run: yum update -y; yum install -y binutils findutils git pciutils sudo tar wget which
- name: Checkout the Repository
uses: actions/checkout@v4
with:
submodules: true
- name: Free Disk Space on Host
run: . $PRELUDE; free_disk_space_on_host
- name: Display System Info
run: . $PRELUDE; print_system_info
- name: Display GPU Info
run: . $PRELUDE; print_gpu_info
- name: Setup Miniconda
run: . $PRELUDE; setup_miniconda $HOME/miniconda
- name: Create Conda Environment
run: . $PRELUDE; create_conda_environment $BUILD_ENV ${{ matrix.python-version }}
- name: Install C/C++ Compilers
run: . $PRELUDE; install_cxx_compiler $BUILD_ENV ${{ matrix.compiler }}
- name: Install Build Tools
run: . $PRELUDE; install_build_tools $BUILD_ENV
- name: Install CUDA
run: . $PRELUDE; install_cuda $BUILD_ENV ${{ matrix.cuda-version }}
# Install via PIP to avoid defaulting to the CPU variant if the GPU variant of the day is not ready
- name: Install PyTorch Nightly
run: . $PRELUDE; install_pytorch_pip $BUILD_ENV nightly cuda/${{ matrix.cuda-version }}
- name: Collect PyTorch Environment Info
if: ${{ success() || failure() }}
run: if . $PRELUDE && which conda; then collect_pytorch_env_info $BUILD_ENV; fi
- name: Install cuDNN
run: . $PRELUDE; install_cudnn $BUILD_ENV "$(pwd)/build_only/cudnn" ${{ matrix.cuda-version }}
- name: Prepare FBGEMM_GPU Build
run: . $PRELUDE; cd fbgemm_gpu; prepare_fbgemm_gpu_build $BUILD_ENV
- name: Build FBGEMM_GPU Wheel
run: . $PRELUDE; cd fbgemm_gpu; build_fbgemm_gpu_package $BUILD_ENV nightly genai
- name: Upload Built Wheel as GHA Artifact
# Cannot upgrade to actions/upload-artifact@v4 yet because GLIBC on the instance is too old
uses: actions/upload-artifact@v3
with:
name: fbgemm_gpu_nightly_genai_${{ matrix.host-machine.arch }}_${{ matrix.compiler }}_py${{ matrix.python-version }}_cu${{ matrix.cuda-version }}.whl
path: fbgemm_gpu/dist/*.whl
if-no-files-found: error
# Download the built artifact from GHA, test on GPU, and push to PyPI
test_artifact:
if: ${{ github.repository_owner != 'pytorch' }}
runs-on: ${{ matrix.host-machine.instance }}
container:
image: amazonlinux:2023
options: --user root --privileged --pid=host
volumes:
- /var/run/docker.sock:/var/run/docker.sock
defaults:
run:
shell: bash
env:
PRELUDE: .github/scripts/setup_env.bash
BUILD_ENV: build_binary
BUILD_VARIANT: genai
ENFORCE_CUDA_DEVICE: 0
CUDA_VISIBLE_DEVICES: -1
ADD_LIBCUDA_SYMLINK: 1
strategy:
fail-fast: false
matrix:
host-machine: [
{ arch: x86, instance: "ubuntu-latest" },
]
python-version: [ "3.9", "3.10", "3.11", "3.12" ]
cuda-version: [ "11.8.0", "12.1.1", "12.4.1" ]
# Specify exactly ONE CUDA version for artifact publish
cuda-version-publish: [ "12.4.1" ]
compiler: [ "gcc", "clang" ]
needs: build_artifact
steps:
- name: Setup Build Container
run: yum update -y; yum install -y binutils findutils git pciutils sudo tar wget which
- name: Checkout the Repository
uses: actions/checkout@v4
with:
submodules: true
- name: Free Disk Space on Host
run: . $PRELUDE; free_disk_space_on_host
- name: Download Wheel Artifact from GHA
# Cannot upgrade to actions/download-artifact@v4 yet because GLIBC on the instance is too old
uses: actions/download-artifact@v3
with:
name: fbgemm_gpu_nightly_genai_${{ matrix.host-machine.arch }}_${{ matrix.compiler }}_py${{ matrix.python-version }}_cu${{ matrix.cuda-version }}.whl
- name: Display System Info
run: . $PRELUDE; print_system_info; print_ec2_info
- name: Display GPU Info
run: . $PRELUDE; print_gpu_info
- name: Setup Miniconda
run: . $PRELUDE; setup_miniconda $HOME/miniconda
- name: Create Conda Environment
run: . $PRELUDE; create_conda_environment $BUILD_ENV ${{ matrix.python-version }}
- name: Install C/C++ Compilers for Updated LIBGCC
run: . $PRELUDE; install_cxx_compiler $BUILD_ENV clang
- name: Install CUDA
run: . $PRELUDE; install_cuda $BUILD_ENV ${{ matrix.cuda-version }}
- name: Install PyTorch Nightly
run: . $PRELUDE; install_pytorch_pip $BUILD_ENV nightly cuda/${{ matrix.cuda-version }}
- name: Collect PyTorch Environment Info
if: ${{ success() || failure() }}
run: if . $PRELUDE && which conda; then collect_pytorch_env_info $BUILD_ENV; fi
- name: Prepare FBGEMM_GPU Build
run: . $PRELUDE; cd fbgemm_gpu; prepare_fbgemm_gpu_build $BUILD_ENV
- name: Install FBGEMM_GPU Wheel
run: . $PRELUDE; install_fbgemm_gpu_wheel $BUILD_ENV *.whl
- name: Test with PyTest
timeout-minutes: 30
run: . $PRELUDE; test_all_fbgemm_gpu_modules $BUILD_ENV