lightning/.azure/gpu-tests-lite.yml

# Python package
# Create and test a Python package on multiple Python versions.
# Add steps that analyze code, save the dist with the build record, publish to a PyPI-compatible index, and more:
# https://docs.microsoft.com/azure/devops/pipelines/languages/python

trigger:
  tags:
    include:
      - '*'
  branches:
    include:
      - "master"
      - "release/*"
      - "refs/tags/*"
  paths:
    include:
      - ".azure/gpu-tests-lite.yml"
      - "requirements/lite/**"
      - "src/lightning_lite/**"
      - "tests/tests_lite/**"
      - "tests/tests_pytorch/run_standalone_tests.sh"
      - "tests/tests_lite/run_standalone_tests.sh"  # a symlink to the one above

pr:
  - "master"
  - "release/*"

jobs:
  - job: testing
    # how long to run the job before automatically cancelling
    timeoutInMinutes: "20"
    # how much time to give 'run always even if cancelled tasks' before stopping them
    cancelTimeoutInMinutes: "2"
    pool: azure-jirka-spot
    container:
      image: "pytorchlightning/pytorch_lightning:base-cuda-py3.9-torch1.12-cuda11.6.1"
      # default shm size is 64m. Increase it to avoid:
      # 'Error while creating shared memory: unhandled system error, NCCL version 2.7.8'
      options: "--runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=all --shm-size=512m"
    workspace:
      clean: all

    steps:
    - bash: |
        lspci | egrep 'VGA|3D'
        whereis nvidia
        nvidia-smi
        which python && which pip
        python --version
        pip --version
        pip list
      displayName: 'Image info & NVIDIA'

    - bash: |
        set -e
        TORCH_VERSION=$(python -c "import torch; print(torch.__version__.split('+')[0])")
        CUDA_VERSION_MM=$(python -c "import torch ; print(''.join(map(str, torch.version.cuda.split('.')[:2])))")
        python ./requirements/pytorch/adjust-versions.py requirements/lite/base.txt ${PYTORCH_VERSION}
        pip install -e .[strategies] --find-links https://download.pytorch.org/whl/cu${CUDA_VERSION_MM}/torch_stable.html
        pip install --requirement requirements/pytorch/devel.txt --find-links https://download.pytorch.org/whl/cu${CUDA_VERSION_MM}/torch_stable.html
        pip list
      env:
        PACKAGE_NAME: pytorch
        FREEZE_REQUIREMENTS: 1
      displayName: 'Install dependencies'

    - bash: |
        set -e
        python requirements/collect_env_details.py
        python -c "import torch ; mgpu = torch.cuda.device_count() ; assert mgpu >= 2, f'GPU: {mgpu}'"
      displayName: 'Env details'

    - bash: python -m coverage run --source lightning_lite -m pytest --ignore benchmarks -v --junitxml=$(Build.StagingDirectory)/test-results.xml --durations=50
      env:
        PL_RUN_CUDA_TESTS: "1"
      workingDirectory: tests/tests_lite
      displayName: 'Testing: Lite standard'
      timeoutInMinutes: "10"

    - bash: bash run_standalone_tests.sh
      workingDirectory: tests/tests_lite
      env:
        PL_RUN_CUDA_TESTS: "1"
        PL_STANDALONE_TESTS_SOURCE: "lightning_lite"
      displayName: 'Testing: Lite standalone tests'
      timeoutInMinutes: "10"

    - bash: |
        python -m coverage report
        python -m coverage xml
        python -m coverage html
        python -m codecov --token=$(CODECOV_TOKEN) --commit=$(Build.SourceVersion) --flags=gpu,pytest --name="GPU-coverage" --env=linux,azure
        ls -l
      workingDirectory: tests/tests_lite
      displayName: 'Statistics'

    - task: PublishTestResults@2
      displayName: 'Publish test results'
      inputs:
        testResultsFiles: '$(Build.StagingDirectory)/test-results.xml'
        testRunTitle: '$(Agent.OS) - $(Build.DefinitionName) - Python $(python.version)'
      condition: succeededOrFailed()