Build Wheels(CU124) for Linux #47
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: Build Wheels(CU124) for Linux # Workflow name | |
| on: | |
| workflow_dispatch: # Manual trigger | |
| permissions: | |
| contents: write | |
| jobs: | |
| build_wheels: | |
| name: Build Wheel ${{ matrix.os }} ${{ matrix.pyver }} ${{ matrix.cuda }} ${{ matrix.releasetag == 'wheels' && 'AVX2' || matrix.releasetag }} | |
| runs-on: ubuntu-22.04 | |
| container: nvidia/cuda:12.4.1-cudnn-devel-ubuntu22.04 | |
| strategy: | |
| matrix: # Define the build matrix directly here | |
| os: ["ubuntu-22.04"] | |
| pyver: ["3.10", "3.11", "3.12", "3.13"] # Python versions | |
| cuda: ["12.4.1"] | |
| releasetag: ["AVX2"] # Controls CMAKE_ARGS for CPU features (even in CUDA build) | |
| cudaarch: ["all"] # Controls target CUDA architectures for nvcc | |
| defaults: | |
| run: | |
| shell: bash | |
| env: | |
| CUDAVER: ${{ matrix.cuda }} | |
| AVXVER: ${{ matrix.releasetag }} | |
| CUDAARCHVER: ${{ matrix.cudaarch }} | |
| steps: | |
| - name: Install dependencies | |
| run: | | |
| apt update | |
| apt install -y build-essential cmake curl git libgomp1 libcurl4-openssl-dev | |
| - uses: actions/checkout@v4 # Checkout code | |
| with: | |
| submodules: "recursive" | |
| # from astral-sh/setup-uv | |
| - name: Install the latest version of uv and set the python version | |
| uses: astral-sh/setup-uv@v6 | |
| with: | |
| python-version: ${{ matrix.pyver }} | |
| activate-environment: true | |
| enable-cache: true | |
| - run: nvcc -V | |
| - name: Build Wheel With Cmake # Main build step: configures and builds the wheel | |
| env: | |
| LD_LIBRARY_PATH: "/usr/local/cuda/lib64:/usr/local/cuda/compat:/usr/lib/x86_64-linux-gnu:${LD_LIBRARY_PATH}" | |
| VERBOSE: 1 # Enable verbose build output | |
| CUDA_HOME: "/usr/local/cuda/" # Set CUDA_HOME | |
| CUDA_PATH: "${PATH}" | |
| CUDA_TOOLKIT_ROOT_DIR: "/usr/local/cuda/" # Set CUDA_TOOLKIT_ROOT_DIR | |
| run: | | |
| echo "VERBOSE=1" >> $GITHUB_ENV # Enable verbose build output for troubleshooting | |
| find /usr/ -name 'libcuda.so.*' | |
| echo $LD_LIBRARY_PATH | |
| # Add project-specific and feature flags | |
| CMAKE_ARGS="-DGGML_CUDA=on -DCMAKE_CUDA_ARCHITECTURES='70-real;75-real;80-real;86-real;89-real'" | |
| CMAKE_ARGS="-DGGML_CUDA_FORCE_MMQ=ON ${CMAKE_ARGS}" | |
| CMAKE_ARGS="${CMAKE_ARGS} -DLLAMA_CURL=ON" | |
| if [ "${AVXVER}" = "AVX" ]; then | |
| CMAKE_ARGS="${CMAKE_ARGS} -DGGML_AVX2=off -DGGML_FMA=off -DGGML_F16C=off" | |
| fi | |
| if [ "${AVXVER}" = "AVX2" ]; then | |
| CMAKE_ARGS="${CMAKE_ARGS} -DGGML_AVX2=on -DGGML_FMA=off -DGGML_F16C=off" | |
| fi | |
| # if [ "${AVXVER}" = "AVX512" ]; then | |
| # CMAKE_ARGS="${CMAKE_ARGS} -DGGML_AVX512=on" | |
| # fi | |
| # if [ "${AVXVER}" = "basic" ]; then | |
| # CMAKE_ARGS = "${CMAKE_ARGS} -DGGML_AVX=off -DGGML_AVX2=off -DGGML_FMA=off -DGGML_F16C=off" | |
| # fi | |
| # Export CMAKE_ARGS environment variable so the python -m build command can use it | |
| echo ${CMAKE_ARGS} | |
| echo "CMAKE_ARGS=${CMAKE_ARGS}" >> $GITHUB_ENV | |
| # Run the Python build command to generate the wheel | |
| uv pip install build setuptools wheel packaging | |
| # uv pip install -U torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu126 | |
| CMAKE_ARGS=${CMAKE_ARGS} uv build --wheel | |
| # --- Post-build steps to get info for release tag --- | |
| # Find the generated wheel file in the 'dist' directory using bash | |
| # Assumes only one wheel is generated per build configuration run | |
| wheel_file=$(ls dist/*.whl | head -n 1) | |
| # Extract the package version (e.g., 1.2.3) from the wheel filename | |
| # Filename format is typically: package_name-version-tag-specificators.whl | |
| # Using basename and cut to split by '-' and get the second field | |
| tag_ver=$(basename "$wheel_file" | cut -d'-' -f 2) | |
| echo "TAG_VERSION=$tag_ver" >> $GITHUB_ENV # Store version in env for release step | |
| # Extract the short CUDA version (e.g., 124) from the full version (e.g., 12.4.1) from the matrix variable | |
| cuda_ver_short=$(echo "${CUDAVER}" | cut -d'.' -f 1,2 | sed 's/\.//g') | |
| echo "CUDA_VERSION=$cuda_ver_short" >> $GITHUB_ENV # Store short CUDA version in env | |
| - name: Get Current Date # Step to get current date for the release tag | |
| id: get-date | |
| run: | | |
| # Get date in YYYYMMDD format using bash date command | |
| currentDate=$(date +%Y%m%d) | |
| # Store the date in environment variable for the release step | |
| echo "BUILD_DATE=$currentDate" >> $GITHUB_ENV | |
| - uses: softprops/[email protected] # Action to create a GitHub Release | |
| with: | |
| files: dist/* # Upload the generated wheel files from the dist directory | |
| # Define the release tag name using the collected environment variables | |
| # Format: v<package_version>-cu<short_cuda_version>-<avx_tag>-linux-<build_date> | |
| tag_name: v${{ env.TAG_VERSION }}-cu${{ env.CUDA_VERSION }}-${{ env.AVXVER }}-linux-${{ env.BUILD_DATE }} # Release tag format for Linux | |
| # Note: This action will create a new release tag if it doesn't exist, | |
| # or upload assets to an existing tag. Be mindful of potential tag name conflicts. | |
| env: | |
| GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} # Use the secret provided by GitHub Actions for authentication |