unit-tests.yaml

name: Unit Tests

on:
  pull_request:
    branches:
    - main
    - release/*

jobs:
  unit-tests:
    name: Build and test with torch-${{ matrix.torch }}
    runs-on: [self-hosted, linux, x64, gpu]
    strategy:
      matrix:
        include:
        # 1.14.0a0+410ce96
        - torch: "1.14"
          nvcr: 22.12-py3
          dir: torch1
        # 2.1.0a0+32f93b1
        - torch: "2.1"
          nvcr: 23.10-py3
          dir: torch2
    container:
      image: nvcr.io/nvidia/pytorch:${{ matrix.nvcr }}
      options: --privileged --ipc=host --gpus=all
    steps:
    - name: Checkout msamp
      uses: actions/checkout@v2
      with:
          submodules: true
          path: ${{ matrix.dir }}
    - name: Install MSCCL
      run: |
        cd ${{ matrix.dir }}/third_party/msccl
        make -j src.build NVCC_GENCODE="\
          -gencode=arch=compute_70,code=sm_70 \
          -gencode=arch=compute_80,code=sm_80 \
          -gencode=arch=compute_90,code=sm_90"
        make install
    - name: Install dependencies
      run: |
        export LD_LIBRARY_PATH="/usr/local/lib:$LD_LIBRARY_PATH"
        export DEBIAN_FRONTEND=noninteractive
        python3 -m pip install --upgrade pip
        apt-get update && apt-get install -y python3-mpi4py
        cd ${{ matrix.dir }}/
        python3 -m pip install .[test]
        make postinstall
    - name: Run code lint
      run: |
        cd ${{ matrix.dir }}/
        python3 setup.py lint
    - name: Run unit tests
      run: |
        export LD_LIBRARY_PATH="/usr/local/lib:$LD_LIBRARY_PATH"
        export LD_PRELOAD="/usr/local/lib/libmsamp_dist.so:/usr/local/lib/libnccl.so:${LD_PRELOAD}"
        cd ${{ matrix.dir }}/
        python3 setup.py test
    # - name: Report coverage results
    #   run: |
    #     bash <(curl -s https://codecov.io/bash)