test_cli_cuda_pytorch_multi_gpu.yaml

name: CLI CUDA Pytorch Multi-GPU Tests

on:
  workflow_dispatch:
  push:
    branches:
      - main
    paths:
      - .github/workflows/test_cli_cuda_pytorch_multi_gpu.yaml
      - "optimum_benchmark/**"
      - "docker/**"
      - "tests/**"
      - "setup.py"
  pull_request:
    branches:
      - main
    paths:
      - .github/workflows/test_cli_cuda_pytorch_multi_gpu.yaml
      - "optimum_benchmark/**"
      - "docker/**"
      - "tests/**"
      - "setup.py"

concurrency:
  cancel-in-progress: true
  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}

env:
  IMAGE: ghcr.io/huggingface/optimum-benchmark:latest-cuda

jobs:
  run_cli_cuda_pytorch_multi_gpu_tests:
    runs-on: [multi-gpu, nvidia-gpu, 4-a10, ci]

    steps:
      - name: Checkout
        uses: actions/checkout@v4

      - name: Run tests
        uses: addnab/docker-run-action@v3
        with:
          image: ${{ env.IMAGE }}
          options: |
            --rm
            --gpus all
            --shm-size 64G
            --env MKL_THREADING_LAYER=GNU
            --volume ${{ github.workspace }}:/workspace
            --workdir /workspace
          run: |
            pip install -e .[testing,diffusers,timm,peft,deepspeed]
            pytest -x -s -k "cli and cuda and pytorch and (dp or ddp or device_map or deepspeed)"