ci.yml

name: Olive Recipes CI
permissions:
  contents: read

on:
  schedule:
    - cron: '0 15 * * 5'
  push:
    branches:
      - main
  workflow_dispatch:
    inputs:
      test_scope:
        type: choice
        description: 'Which tests to run'
        options:
          - all
          - linux-cpu-only
          - linux-gpu-only
          - windows-cpu-only
        default: all

env:
  PYTHON_VERSION: "3.10"

jobs:
  linux-cpu:
    name: Linux CPU - ${{ matrix.test.name }}
    if: |
      github.event.inputs.test_scope == 'all' ||
      github.event.inputs.test_scope == 'linux-cpu-only' ||
      github.event.inputs.test_scope == ''
    runs-on: ubuntu-latest
    timeout-minutes: 120
    strategy:
      fail-fast: false
      matrix:
        test:
          - name: "PTQ"
            path: "intel-bert-base-uncased-mrpc/oci/cpu"
            config: "ptq.json"
          - name: "INC Smooth Quant"
            path: "intel-bert-base-uncased-mrpc/oci/cpu"
            config: "inc_smooth_quant.json"
          - name: "ResNet-50 Session Param Tuning"
            path: "microsoft-resnet-50/olive"
            config: "resnet_cpu.json"

    steps:
      - uses: actions/checkout@v4

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

      - name: Install Olive
        run: pip install git+https://github.com/microsoft/Olive.git@main

      - name: Install optimum-onnx for HuggingFace export
        run: pip install optimum-onnx

      - name: Install requirements
        working-directory: ${{ matrix.test.path }}
        run: |
          if [ -f requirements.txt ]; then pip install -r requirements.txt; fi

      - name: Run Olive workflow and verify output
        working-directory: ${{ matrix.test.path }}
        env:
          HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
        run: python ${{ github.workspace }}/.github/scripts/verify_model.py ${{ matrix.test.config }}

  # ========================================
  # GPU TESTS TEMPORARILY DISABLED and Will re-enable once pool hardware is upgraded
  # ========================================

  # linux-gpu:
  #   name: Linux GPU - ${{ matrix.name }}
  #   if: |
  #     github.event.inputs.test_scope == 'all' ||
  #     github.event.inputs.test_scope == 'linux-gpu-only' ||
  #     github.event.inputs.test_scope == ''
  #   runs-on: ["self-hosted", "1ES.Pool=olive-github-ubuntu2204-cuda-A10"]
  #   timeout-minutes: 180
  #   strategy:
  #     fail-fast: false
  #     max-parallel: 1
  #     matrix:
  #       include:
  #         - name: "Dora"
  #           path: "meta-llama-Llama-3.2-1B-Instruct/olive"
  #           config: "dora.json"
  #         - name: "HQQ"
  #           path: "meta-llama-Llama-3.2-1B-Instruct/olive"
  #           config: "hqq.json"
  #         - name: "Lmeval"
  #           path: "meta-llama-Llama-3.2-1B-Instruct/olive"
  #           config: "lmeval.json"
  #         - name: "Lmeval-ONNX"
  #           path: "meta-llama-Llama-3.2-1B-Instruct/olive"
  #           config: "lmeval_onnx.json"
  #         - name: "Loha"
  #           path: "meta-llama-Llama-3.2-1B-Instruct/olive"
  #           config: "loha.json"
  #         - name: "Lokr"
  #           path: "meta-llama-Llama-3.2-1B-Instruct/olive"
  #           config: "lokr.json"
  #         - name: "QLoRA"
  #           path: "meta-llama-Llama-3.2-1B-Instruct/olive"
  #           config: "qlora.json"
  #         - name: "RTN"
  #           path: "meta-llama-Llama-3.2-1B-Instruct/olive"
  #           config: "rtn.json"
  #
  #   steps:
  #     - name: DEBUG - Print system and GPU information
  #       run: |
  #         echo "=========================================="
  #         echo "DEBUG: This step is running!"
  #         echo "=========================================="
  #         uname -a || true
  #         echo "--- Attempting to get Azure VM metadata ---"
  #         curl -H Metadata:true --noproxy "*" "http://169.254.169.254/metadata/instance?api-version=2021-02-01" 2>/dev/null | jq '.' || echo "Metadata not available"
  #         echo "--- GPU Information ---"
  #         nvidia-smi --query-gpu=name,compute_cap,driver_version,memory.total --format=csv || echo "nvidia-smi failed"
  #         nvidia-smi || echo "nvidia-smi not found"
  #         echo "=========================================="
  #         echo "DEBUG: Step completed!"
  #         echo "=========================================="
  #
  #     - uses: actions/checkout@v4
  #
  #     - uses: actions/setup-python@v5
  #       with:
  #         python-version: ${{ env.PYTHON_VERSION }}
  #
  #     - name: Install Olive
  #       run: pip install git+https://github.com/microsoft/Olive.git@main
  #
  #     - name: Install ONNX Runtime GPU
  #       run: pip install onnxruntime-gpu
  #
  #     - name: Install optimum-onnx for HuggingFace export
  #       run: pip install optimum-onnx
  #
  #     - name: Install requirements
  #       working-directory: ${{ matrix.path }}
  #       run: |
  #         if [ -f requirements.txt ]; then
  #           pip install -r requirements.txt
  #         fi
  #
  #     - name: Run Olive workflow and verify output
  #       working-directory: ${{ matrix.path }}
  #       env:
  #         HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
  #       run: python ${{ github.workspace }}/.github/scripts/verify_model.py ${{ matrix.config }}
  #
  #     - name: Cleanup GPU
  #       if: always()
  #       run: |
  #         nvidia-smi
  #         python -c "import torch; torch.cuda.empty_cache()" || true
  #     - name: Upload artifacts
  #       if: always()
  #       uses: actions/upload-artifact@v4
  #       with:
  #         name: linux-gpu-${{ matrix.name }}-output
  #         path: ${{ matrix.path }}/models/
  #         retention-days: 7

  windows-cpu:
    name: Windows CPU - ${{ matrix.test.name }}
    if: |
      github.event.inputs.test_scope == 'all' ||
      github.event.inputs.test_scope == 'windows-cpu-only' ||
      github.event.inputs.test_scope == ''
    runs-on: windows-latest
    timeout-minutes: 120
    strategy:
      fail-fast: false
      matrix:
        test:
          - name: "PTQ"
            path: "intel-bert-base-uncased-mrpc/oci/cpu"
            config: "ptq.json"
          # Temporarily disabled - has dtype issues
          # - name: "INC Smooth Quant"
          #   path: "intel-bert-base-uncased-mrpc/oci/cpu"
          #   config: "inc_smooth_quant.json"

    steps:
      - uses: actions/checkout@v4

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

      - name: Install Olive
        run: pip install git+https://github.com/microsoft/Olive.git@main

      - name: Install optimum-onnx for HuggingFace export
        run: pip install optimum-onnx

      - name: Install requirements
        working-directory: ${{ matrix.test.path }}
        shell: pwsh
        run: |
          if (Test-Path requirements.txt) { pip install -r requirements.txt }

      - name: Run Olive workflow and verify output
        working-directory: ${{ matrix.test.path }}
        shell: pwsh
        env:
          HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
          PYTHONIOENCODING: utf-8
        run: python ${{ github.workspace }}/.github/scripts/verify_model.py ${{ matrix.test.config }}