hf_text-generation-inference/.github/workflows/build.yaml

name: Build and push docker image to internal registry

on:
  workflow_call:
    inputs:
      hardware:
        type: string
        description: Hardware
          # options:
          # - cuda
          # - rocm
          # - xpu
        required: true
      release-tests:
        description: "Run release integration tests"
        required: true
        default: false
        type: boolean

jobs:
  build-and-push:
    outputs:
      docker_image: ${{ steps.final.outputs.docker_image }}
      docker_devices: ${{ steps.final.outputs.docker_devices }}
      docker_volume: ${{ steps.final.outputs.docker_volume}}
      runs_on: ${{ steps.final.outputs.runs_on }}
      label: ${{ steps.final.outputs.label }}
    concurrency:
      group: ${{ github.workflow }}-build-and-push-image-${{ inputs.hardware }}-${{ github.head_ref || github.run_id }}
      cancel-in-progress: true
    # TODO see with @Glegendre to get CPU runner here instead
    runs-on: [self-hosted, nvidia-gpu , multi-gpu, 4-a10, ci]
    permissions:
      contents: write
      packages: write
      # This is used to complete the identity challenge
      # with sigstore/fulcio when running outside of PRs.
      id-token: write
      security-events: write

    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
      - name: Inject slug/short variables
        uses: rlespinasse/github-slug-action@v4.4.1

      - name: Construct harware variables
        shell: bash
        run: |
          case ${{ inputs.hardware }} in
            cuda)
                export dockerfile="Dockerfile"
                export label_extension=""
                export docker_devices=""
                export runs_on="nvidia-gpu"
                ;;
            rocm)
                export dockerfile="Dockerfile_amd"
                export label_extension="-rocm"
                export docker_devices="/dev/kfd,/dev/dri"
                export runs_on="amd-gpu-tgi"
                ;;
            xpu)
                export dockerfile="Dockerfile_intel"
                export label_extension="-intel"
                export docker_devices=""
                export runs_on="ubuntu-latest"
                ;;
          esac
          echo $dockerfile
          echo "Dockerfile=${dockerfile}"
          echo $label_extension
          echo $docker_devices
          echo $runs_on
          echo "DOCKERFILE=${dockerfile}" >> $GITHUB_ENV
          echo "LABEL=${label_extension}" >> $GITHUB_ENV
          echo "DOCKER_DEVICES=${docker_devices}" >> $GITHUB_ENV
          echo "RUNS_ON=${runs_on}" >> $GITHUB_ENV

      - name: Initialize Docker Buildx
        uses: docker/setup-buildx-action@v3
        with:
          install: true
          config-inline: |
            [registry."docker.io"]
              mirrors = ["registry.github-runners.huggingface.tech"]

      - name: Login to GitHub Container Registry
        if: github.event_name != 'pull_request'
        uses: docker/login-action@v3
        with:
          registry: ghcr.io
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}

      - name: Login to Azure Container Registry
        if: github.event_name != 'pull_request'
        uses: docker/login-action@v3
        with:
          username: ${{ secrets.AZURE_DOCKER_USERNAME }}
          password: ${{ secrets.AZURE_DOCKER_PASSWORD }}
          registry: db4c2190dd824d1f950f5d1555fbadf0.azurecr.io

      # If pull request
      - name: Extract metadata (tags, labels) for Docker
        if: ${{ github.event_name == 'pull_request' }}
        id: meta-pr
        uses: docker/metadata-action@v5
        with:
          images: |
            registry-push.github-runners.huggingface.tech/api-inference/community/text-generation-inference
          tags: |
            type=raw,value=sha-${{ env.GITHUB_SHA_SHORT }}${{ env.LABEL }}
      # If main, release or tag
      - name: Extract metadata (tags, labels) for Docker
        if: ${{ github.event_name != 'pull_request' }}
        id: meta
        uses: docker/metadata-action@v4.3.0
        with:
          flavor: |
            latest=auto
          images: |
            registry-push.github-runners.huggingface.tech/api-inference/community/text-generation-inference
            ghcr.io/huggingface/text-generation-inference
            db4c2190dd824d1f950f5d1555fbadf0.azurecr.io/text-generation-inference
          tags: |
            type=semver,pattern={{version}}${{ env.LABEL }}
            type=semver,pattern={{major}}.{{minor}}${{ env.LABEL }}
            type=raw,value=latest${{ env.LABEL }},enable=${{ github.ref == format('refs/heads/{0}', github.event.repository.default_branch) }}
            type=raw,value=sha-${{ env.GITHUB_SHA_SHORT }}${{ env.LABEL }}
      - name: Build and push Docker image
        id: build-and-push
        uses: docker/build-push-action@v4
        with:
          context: .
          file: ${{ env.DOCKERFILE }}
          push: true
          platforms: 'linux/amd64'
          build-args: |
            GIT_SHA=${{ env.GITHUB_SHA }}
            DOCKER_LABEL=sha-${{ env.GITHUB_SHA_SHORT }}${{ env.LABEL }}
          tags: ${{ steps.meta.outputs.tags || steps.meta-pr.outputs.tags }}
          labels: ${{ steps.meta.outputs.labels || steps.meta-pr.outputs.labels }}
          cache-from: type=registry,ref=registry-push.github-runners.huggingface.tech/api-inference/community/text-generation-inference:cache${{ env.LABEL }},mode=min
          cache-to: type=registry,ref=registry-push.github-runners.huggingface.tech/api-inference/community/text-generation-inference:cache${{ env.LABEL }},mode=min
      - name: Final
        id: final
        run: |
          echo "docker_image=registry-push.github-runners.huggingface.tech/api-inference/community/text-generation-inference:sha-${{ env.GITHUB_SHA_SHORT}}${{ env.LABEL }}" >> "$GITHUB_OUTPUT"
          echo "docker_devices=${{ env.DOCKER_DEVICES }}" >> "$GITHUB_OUTPUT"
          echo "runs_on=${{ env.RUNS_ON }}" >> "$GITHUB_OUTPUT"
          echo "label=${{ env.LABEL }}" >> "$GITHUB_OUTPUT"

          if [[ ${{ inputs.hardware }} == "rocm" ]]
          then
            echo "docker_volume=/data/cache/.cache/huggingface/hub" >> "$GITHUB_OUTPUT"
          else
            echo "docker_volume=/mnt/cache" >> "$GITHUB_OUTPUT"
          fi

  prepare_integration_tests:
    runs-on: ["self-hosted", "${{ needs.build-and-push.outputs.runs_on }}", "multi-gpu"]
    needs: [build-and-push]
    concurrency:
      group: ${{ github.workflow }}-${{ github.job }}-${{ needs.build-and-push.outputs.label }}-${{ github.head_ref || github.run_id }}
      cancel-in-progress: true
    if: needs.build-and-push.outputs.runs_on != 'ubuntu-latest'
    # Ideally, we would use the image from registry.internal.huggingface.tech but we can not login to the private registry outside of tailscale,
    # and even adding a previous job with tailscale login still results in `Docker login for 'registry.internal.huggingface.tech' failed with exit code 1`.
    container:
      image: ${{ needs.build-and-push.outputs.docker_image }}
      options: --shm-size "16gb" --ipc host -v ${{ needs.build-and-push.outputs.docker_volume }}:/data
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4

      - name: Clean Hugging Face cache
        shell: bash
        run: |
          if [[ ${{ inputs.hardware }} == "rocm" ]]
          then
            echo "pwd:"
            pwd
            echo "ls:"
            ls

            python3 integration-tests/clean_cache_and_download.py --token ${{ secrets.HF_TOKEN }} --cache-dir /data

            # Avoid permissions issues in the next step not run within docker (File was unable to be removed Error: EACCES).
            if [[ $PWD == *"text-generation-inference"* ]]; then
              rm -rf -- ..?* .[!.]* *
            fi
          fi

  integration_tests:
    concurrency:
      group: ${{ github.workflow }}-${{ github.job }}-${{ needs.build-and-push.outputs.label }}-${{ github.head_ref || github.run_id }}
      cancel-in-progress: true
    needs: [build-and-push, prepare_integration_tests]
    runs-on: ["self-hosted", "${{ needs.build-and-push.outputs.runs_on }}", "multi-gpu"]
    if: needs.build-and-push.outputs.runs_on != 'ubuntu-latest'
    env:
      PYTEST_FLAGS: ${{ (startsWith(github.ref, 'refs/tags/') || github.ref == 'refs/heads/main' || inputs.release-tests == true) && '--release' || '' }}
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4

      - name: Inject slug/short variables
        uses: rlespinasse/github-slug-action@v4.4.1

      - name: Set up Python
        uses: actions/setup-python@v4
        with:
          python-version: "3.10"

      - name: Install
        run: |
          make install-integration-tests

      - name: Run tests
        run: |
          export DOCKER_DEVICES=${{ needs.build-and-push.outputs.docker_devices }}
          export HF_TOKEN=${{ secrets.HF_TOKEN }}

          export DOCKER_IMAGE=${{ needs.build-and-push.outputs.docker_image }}
          echo "DOCKER_IMAGE:"
          echo $DOCKER_IMAGE

          export SYSTEM=${{ inputs.hardware }}
          echo "SYSTEM:"
          echo $SYSTEM

          export DOCKER_VOLUME=${{ needs.build-and-push.outputs.docker_volume }}
          echo "DOCKER_VOLUME:"
          echo $DOCKER_VOLUME

          # TunableOp warmup is rather slow, do it only for a few seqlens.
          if [[ ${{ inputs.hardware }} == "rocm" ]]
          then
            PYTORCH_TUNABLEOP_SEQLENS=2,4
          fi

          pytest -s -vvvvv integration-tests ${PYTEST_FLAGS}