exllamav2/.github/workflows/build-wheels.yml

name: Build Wheels

on: workflow_dispatch

jobs:
  build_wheels:
    name: ${{ matrix.os }} Python ${{ matrix.pyver }} CUDA ${{ matrix.cuda }}
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
        os: [ubuntu-20.04, windows-latest]
        pyver: ["3.8", "3.9", "3.10", "3.11"]
        cuda: ["11.7.0", "11.8.0", "12.1.1"]
    defaults:
      run:
        shell: pwsh
    env:
      CUDAVER: ${{ matrix.cuda }}
      PYVER: ${{ matrix.pyver }}

    steps:
      - name: Free Disk Space
        uses: jlumbroso/free-disk-space@v1.2.0
        if: runner.os == 'Linux'
        with:
          tool-cache: false
          android: true
          dotnet: true
          haskell: true
          large-packages: false
          swap-storage: false
          
      - uses: actions/checkout@v3
      - uses: actions/setup-python@v3
        with:
          python-version: ${{ matrix.pyver }}
        
      - name: Setup Mamba
        uses: conda-incubator/setup-miniconda@v2.2.0
        with:
          activate-environment: "build"
          python-version: ${{ matrix.pyver }}
          miniforge-variant: Mambaforge
          miniforge-version: latest
          use-mamba: true
          add-pip-as-python-dependency: true
          auto-activate-base: false
          
      - name: Install Dependencies
        run: |
          $cudaVersion = $env:CUDAVER
          $cudaVersionPytorch = $env:CUDAVER.Remove($env:CUDAVER.LastIndexOf('.')).Replace('.','')
          
          $cudaChannels = ''
          $cudaNum = [int]$cudaVersion.substring($cudaVersion.LastIndexOf('.')+1)
          while ($cudaNum -ge 0) { $cudaChannels += '-c nvidia/label/cuda-' + $cudaVersion.Remove($cudaVersion.LastIndexOf('.')+1) + $cudaNum + ' '; $cudaNum-- }
          mamba install -y 'cuda' $cudaChannels.TrimEnd().Split()
          
          if (!(mamba list cuda)[-1].contains('cuda')) {sleep -s 10; mamba install -y 'cuda' $cudaChannels.TrimEnd().Split()}
          if (!(mamba list cuda)[-1].contains('cuda')) {throw 'CUDA Toolkit failed to install!'}
          
          if ([version]$env:CUDAVER -lt [version]'11.8.0') {$torchver = "torch==2.0.1"} else {$torchver = "torch==2.1.0"}
          python -m pip install $torchver --index-url https://download.pytorch.org/whl/cu$cudaVersionPytorch
          
          python -m pip install build wheel safetensors sentencepiece ninja
        
      - name: Build Wheel
        run: |
          $env:CUDA_PATH = $env:CONDA_PREFIX
          $env:CUDA_HOME = $env:CONDA_PREFIX
          
          $cudaVersion = $env:CUDAVER
          $cudaVersionPytorch = $env:CUDAVER.Remove($env:CUDAVER.LastIndexOf('.')).Replace('.','')
          $BUILDTAG = "+cu$cudaVersionPytorch"
          
          if ($IsLinux) {$env:LD_LIBRARY_PATH = $env:CONDA_PREFIX + '/lib:' + $env:LD_LIBRARY_PATH}
          $env:TORCH_CUDA_ARCH_LIST = if ([version]$env:CUDAVER -lt [version]'11.8') {'6.0 6.1 7.0 7.5 8.0 8.6+PTX'} else {'6.0 6.1 7.0 7.5 8.0 8.6 8.9 9.0+PTX'}
          python -m build -n --wheel -C--build-option=egg_info "-C--build-option=--tag-build=$BUILDTAG"
          
          if ($IsLinux -and $env:PYVER -eq '3.11' -and $env:CUDAVER -eq '11.8.0') {$env:EXLLAMA_NOCOMPILE=1; python -m build -n}
        
      - uses: actions/upload-artifact@v3
        with:
          name: 'wheels'
          path: ./dist/*
          
  build_rocm:
    name: Build ROCm Wheels & Release
    needs: build_wheels
    uses: ./.github/workflows/build-wheels-rocm.yml