test_cli_cuda_torch_ort.yaml

name: CLI CUDA Torch-ORT Tests

on:
  workflow_dispatch:
  push:
    branches:
      - main
  pull_request:
    branches:
      - main
    types:
      - opened
      - reopened
      - synchronize
      - labeled
      - unlabeled

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

jobs:
  run_cli_cuda_torch_ort_single_gpu_tests:
    if: ${{
      (github.event_name == 'push') ||
      (github.event_name == 'workflow_dispatch') ||
      contains( github.event.pull_request.labels.*.name, 'cli') ||
      contains( github.event.pull_request.labels.*.name, 'cuda') ||
      contains( github.event.pull_request.labels.*.name, 'torch_ort') ||
      contains( github.event.pull_request.labels.*.name, 'single_gpu') ||
      contains( github.event.pull_request.labels.*.name, 'cli_cuda_torch_or_single_gpu')
      }}

    runs-on:
      group: aws-g5-4xlarge-plus

    container:
      image: ghcr.io/huggingface/optimum-benchmark:latest-cuda-ort
      options: --ipc host --gpus all

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

      - name: Install dependencies
        run: |
          pip install -e .[testing,torch-ort,peft]
          pip install optimum@git+https://github.com/huggingface/optimum.git

      - name: Run tests
        run: |
          pytest tests/test_cli.py -x -s -k "cli and cuda and torch_ort and not (dp or ddp or device_map) and not (peft)"

  run_cli_cuda_torch_ort_multi_gpu_tests:
    if: ${{
      (github.event_name == 'push') ||
      (github.event_name == 'workflow_dispatch') ||
      contains( github.event.pull_request.labels.*.name, 'cli') ||
      contains( github.event.pull_request.labels.*.name, 'cuda') ||
      contains( github.event.pull_request.labels.*.name, 'torch_ort') ||
      contains( github.event.pull_request.labels.*.name, 'multi_gpu') ||
      contains( github.event.pull_request.labels.*.name, 'cli_cuda_torch_ort_multi_gpu')
      }}

    runs-on:
      group: aws-g5-12xlarge-plus

    container:
      image: ghcr.io/huggingface/optimum-benchmark:latest-cuda-ort
      options: --ipc host --gpus all

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

      - name: Install dependencies
        run: |
          pip install -e .[testing,torch-ort,peft] 
          pip install optimum@git+https://github.com/huggingface/optimum.git

      - name: Run tests
        run: |
          pytest tests/test_cli.py -x -s -k "cli and cuda and torch_ort and (dp or ddp or device_map)"