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

name: Build Wheels & Release

on:
  workflow_dispatch:
    inputs:
      pypi:
        description: 'Upload wheels to PyPI? 1 = yes, 0 = no'
        default: '0'
        required: true
        type: string


permissions:
  contents: write

jobs:
  build_wheels:
    name: ${{ matrix.os }} Python ${{ matrix.pyver }} CUDA ${{ matrix.cuda }}
    runs-on: ${{ matrix.os }}
    strategy:
      matrix:
        os: [ubuntu-latest, 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@v2.3.0
        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==2.2.0"}
          python -m pip install $torchver --index-url https://download.pytorch.org/whl/cu$cudaVersionPytorch
          
          python -m pip install --upgrade setuptools==65.5.1
          python -m pip install --upgrade build wheel safetensors sentencepiece ninja

      - name: Build Wheel
        id: build-wheel
        run: |
          $versionString = Get-Content $(Join-Path 'exllamav2' 'version.py') -raw
          if ($versionString -match '__version__ = "(\d+\.(?:\d+\.?(?:dev\d+)?)*)"')
          {
              Write-Output $('::notice file=build-wheels-release.yml,line=83,title=Package Version::Detected package version is: {0}' -f $Matches[1])
              Write-Output "PACKAGE_VERSION=$($Matches[1])" >> "$env:GITHUB_OUTPUT"
          } else {
              Write-Output '::error file=build-wheels-release.yml,line=86::Could not parse version from exllamav2/version.py! You must upload wheels manually!'
              Write-Output "PACKAGE_VERSION=None" >> "$env:GITHUB_OUTPUT"
          }
          
          $env:CUDA_PATH = $env:CONDA_PREFIX
          $env:CUDA_HOME = $env:CONDA_PREFIX
          if ($IsLinux) {$env:LD_LIBRARY_PATH = $env:CONDA_PREFIX + '/lib:' + $env:LD_LIBRARY_PATH}
          
          $cudaVersion = $env:CUDAVER
          $cudaVersionPytorch = $env:CUDAVER.Remove($env:CUDAVER.LastIndexOf('.')).Replace('.','')
          $BUILDTAG = "+cu$cudaVersionPytorch"
          
          $env:TORCH_CUDA_ARCH_LIST = if ([version]$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"
        
      - uses: actions/upload-artifact@v3
        with:
          name: 'wheels'
          path: ./dist/*
          
      - name: Upload files to a GitHub release
        if: steps.build-wheel.outputs.PACKAGE_VERSION != 'None'
        uses: svenstaro/upload-release-action@2.6.1
        with:
          file: ./dist/*.whl
          tag: ${{ format('v{0}', steps.build-wheel.outputs.PACKAGE_VERSION) }}
          file_glob: true
          overwrite: true
          release_name: ${{ steps.build-wheel.outputs.PACKAGE_VERSION }}
          
  build_rocm:
    name: Build ROCm Wheels & Release
    needs: build_wheels
    uses: ./.github/workflows/build-wheels-release-rocm.yml
    with:
      pypi: '0'
      
  # publish-wheels-to-pypi:
  #   name: Publish Python distribution to PyPI
  #   if: inputs.pypi == '1'
  #   needs: ['build_wheels','build_rocm']
  #   runs-on: ubuntu-latest
  #   
  #   environment:
  #     name: pypi
  #     url: https://pypi.org/p/exllamav2
  #   permissions:
  #     id-token: write  # IMPORTANT: mandatory for trusted publishing
  #     
  #   steps:
  #     - name: Download all the wheels
  #       uses: actions/download-artifact@v3
  #       with:
  #         name: wheels
  #         path: dist/
  #     - name: Publish distribution to PyPI
  #       uses: pypa/gh-action-pypi-publish@v1.8.10
        
  build_sdist:
    name: Build sdist
    needs: ['build_wheels','build_rocm'] # publish-wheels-to-pypi
    runs-on: ubuntu-20.04
    defaults:
      run:
        shell: pwsh

    steps:
      - uses: actions/checkout@v3
      - uses: actions/setup-python@v3
        with:
          python-version: "3.11"
          
      - name: Install Dependencies
        run: |
          python -m pip install torch==2.2.0 --index-url https://download.pytorch.org/whl/cpu
          python -m pip install build wheel ninja
        
      - name: Build Wheel
        id: build-wheel-sdist
        run: |
          $versionString = Get-Content $(Join-Path 'exllamav2' 'version.py') -raw
          if ($versionString -match '__version__ = "(\d+\.(?:\d+\.?(?:dev\d+)?)*)"')
          {
              Write-Output $('::notice file=build-wheels-release.yml,line=166,title=Package Version::Detected package version is: {0}' -f $Matches[1])
              Write-Output "PACKAGE_VERSION=$($Matches[1])" >> "$env:GITHUB_OUTPUT"
          } else {
              Write-Output '::error file=build-wheels-release.yml,line=169::Could not parse version from exllamav2/version.py! You must upload wheels manually!'
              Write-Output "PACKAGE_VERSION=None" >> "$env:GITHUB_OUTPUT"
          }
          
          $env:EXLLAMA_NOCOMPILE=1
          python -m build -n
        
      - uses: actions/upload-artifact@v3
        with:
          name: 'sdist'
          path: ./dist/* # Use ./dist/*.whl to only upload the JIT compile wheel
          
      - name: Upload files to a GitHub release
        if: steps.build-wheel-sdist.outputs.PACKAGE_VERSION != 'None'
        uses: svenstaro/upload-release-action@2.6.1
        with:
          file: ./dist/*.whl
          tag: ${{ format('v{0}', steps.build-wheel-sdist.outputs.PACKAGE_VERSION) }}
          file_glob: true
          overwrite: true
          release_name: ${{ steps.build-wheel-sdist.outputs.PACKAGE_VERSION }}
      
  publish-sdist-to-pypi:
    name: Publish Python distribution to PyPI
    if: inputs.pypi == '1'
    needs: build_sdist
    runs-on: ubuntu-latest
    
    environment:
      name: pypi
      url: https://pypi.org/p/exllamav2
    permissions:
      id-token: write  # IMPORTANT: mandatory for trusted publishing
      
    steps:
      - name: Download the sdist
        uses: actions/download-artifact@v3
        with:
          name: sdist
          path: dist/
      - name: Publish distribution to PyPI
        uses: pypa/gh-action-pypi-publish@v1.8.10