flash-attention-prebuild-wheels-rocm/.github/workflows/test-self-hosted-linux.yml

name: Test self-hosted Linux build

on:
  workflow_dispatch:

jobs:
  # #########################################################
  # Build wheels with self-hosted runner
  # #########################################################
  build_wheels_self_hosted:
    name: Build wheels and Upload (Linux x86_64, self-hosted runner)
    runs-on: self-hosted
    container:
      image: ubuntu:22.04
    defaults:
      run:
        shell: bash
    env:
      DEBIAN_FRONTEND: noninteractive
      TERM: xterm-256color
    timeout-minutes: 2000
    strategy:
      fail-fast: false
      matrix:
        flash-attn-version: ["2.8.0"]
        python-version: ["3.11"]
        torch-version: ["2.7.1"]
        # https://developer.nvidia.com/cuda-toolkit-archive
        cuda-version: ["12.8.1"]
    steps:
      - name: Install tools
        shell: bash
        run: |
          apt-get update && apt-get install -y --no-install-recommends \
            curl \
            ca-certificates \
            sudo \
            software-properties-common \
            wget \
            unzip \
            zip \
            git \
            build-essential \
            gcc \
            g++ \
            clang \
            ninja-build \
            keyboard-configuration

      - name: Install gh
        shell: bash
        run: |
          sudo mkdir -p -m 755 /etc/apt/keyrings
          out=$(mktemp)
          wget -nv -O$out https://cli.github.com/packages/githubcli-archive-keyring.gpg
          cat $out | sudo tee /etc/apt/keyrings/githubcli-archive-keyring.gpg > /dev/null
          sudo chmod go+r /etc/apt/keyrings/githubcli-archive-keyring.gpg
          echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/githubcli-archive-keyring.gpg] https://cli.github.com/packages stable main" | sudo tee /etc/apt/sources.list.d/github-cli.list > /dev/null
          sudo apt update
          sudo apt install gh -y

      - uses: actions/checkout@v4

      - name: Configure Git safe directory
        shell: bash
        run: |
          git config --global --add safe.directory $(pwd)

      - uses: actions/setup-python@v5
        with:
          python-version: ${{ matrix.python-version }}

      - uses: Jimver/cuda-toolkit@master
        with:
          cuda: ${{ matrix.cuda-version }}
          sub-packages: '["nvcc", "toolkit"]'
          method: "network"
          use-github-cache: false
          use-local-cache: false

      - name: Install build dependencies
        shell: bash
        run: |
          sudo apt install -y ninja-build clang
          pip install -U pip setuptools==75.8.0 wheel setuptools packaging psutil

      - name: Set environment variables
        shell: bash
        run: |
          export PATH=/usr/local/nvidia/bin:/usr/local/nvidia/lib64:$PATH
          export LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:$LD_LIBRARY_PATH

      - name: Build wheels
        timeout-minutes: 1200
        shell: bash
        env:
          MAX_JOBS: 4
          NVCC_THREADS: 3
        run: |
          chmod +x build_linux.sh
          ./build_linux.sh ${{ matrix.flash-attn-version }} ${{ matrix.python-version }} ${{ matrix.torch-version }} ${{ matrix.cuda-version }}
          wheel_name=$(basename $(ls flash-attention/dist/*.whl | head -n 1))
          echo "wheel_name=$wheel_name" >> $GITHUB_ENV

      - name: Install Test
        shell: bash
        run: |
          pip install --no-cache-dir --force-reinstall flash-attention/dist/${{ env.wheel_name }}
          python -c "import flash_attn; print(flash_attn.__version__)"

      - name: Clean up
        shell: bash
        if: always()
        run: |
          rm -rf /opt/hostedtoolcache/Python
          rm -rf ~/.cache/pip