mlperf-inference-resnet50.yml

name: 'MLPerf inference resnet50'

on:
  pull_request:
    branches: [ "main", "dev" ]
    paths:
      - '.github/workflows/mlperf-inference-resnet50.yml'
      - '**'
      - '!**.md'

jobs:
  build:

    strategy:
      fail-fast: false
      matrix:
        python-version: ["3.12", "3.11", "3.8"]
        on: [ubuntu-latest, macos-latest, windows-latest]
        backend: [ "onnxruntime", "tf" ]
        implementation: [ "python", "cpp" ]
        exclude:
          - backend: tf
            implementation: cpp
          - on: windows-latest
            implementation: cpp
    runs-on: "${{ matrix.on }}"
    steps:
    - uses: actions/checkout@v4
      with:
        ref: ${{ github.event.pull_request.head.sha }}
    - name: Set up Python ${{ matrix.python-version }}
      uses: actions/setup-python@v3
      with:
        python-version: ${{ matrix.python-version }}
  
    - name: Install dependencies
      run: |
        python -m pip install --upgrade pip
        python -m pip install --ignore-installed --verbose pip setuptools
        python -m pip install .
        mlc pull repo mlcommons@mlperf-automations --branch=dev
  
    - name: Test MLPerf inference ResNet50 on Windows (prebuilt loadgen)
      if: runner.os == 'Windows'
      run: |
        mlc run script --tags=run-mlperf,inference,_submission,_short --submitter="MLCommons" --hw_name=gh_action --model=resnet50 --implementation=${{ matrix.implementation }} --backend=${{ matrix.backend }} --device=cpu --scenario=Offline --test_query_count=100 --target_qps=1 -v --quiet --adr.loadgen.tags=_from-pip --pip_loadgen=yes

    - name: Test MLPerf inference ResNet50 on Unix systems
      if: runner.os != 'Windows'
      run: |
        mlc run script --tags=run-mlperf,inference,_submission,_short --submitter="MLCommons" --hw_name=gh_action --model=resnet50 --implementation=${{ matrix.implementation }} --backend=${{ matrix.backend }} --device=cpu --scenario=Offline --test_query_count=100 --target_qps=1 -v --quiet