cavis/.github/workflows/run-gpu-integration-tests-self-hosted.yml

on:
  workflow_dispatch:
jobs:
  # Wait for up to a minute for previous run to complete, abort if not done by then
  pre-ci:
    runs-on: self-hosted
    timeout-minutes: 1
    steps:
      - name: 'Block Concurrent Executions'
        uses: softprops/turnstyle@v1
        with:
          poll-interval-seconds: 10
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
  linux-x86_64:
    needs: pre-ci
    runs-on: [self-hosted]
    steps:
      - uses: AutoModality/action-clean@v1
      - name: Cancel Previous Runs
        uses: styfle/cancel-workflow-action@0.8.0
        with:
          access_token: ${{ github.token }}
      - uses: ./.github/actions/download-dl4j-test-resources-linux
      - uses: actions/checkout@v2
      - name: Run cpu tests
        shell: bash
        env:
          DEBIAN_FRONTEND: noninteractive
        run: |
          export PATH="/opt/protobuf/bin:/usr/local/cuda-11/bin:$PATH"
          nvcc --version
          mvn --version
          cmake --version
          protoc --version
          export OMP_NUM_THREADS=1
          mkdir -p ${GITHUB_WORKSPACE}/resources
          mkdir -p ${GITHUB_WORKSPACE}/cache
          echo "Running tests for cuda 11.0"
          export PATH="/opt/protobuf/bin:/usr/local/cuda-11.2/bin:$PATH"
          mvn -Dorg.nd4j.strumpf.resource.dirs=${GITHUB_WORKSPACE}/resources -Dorg.nd4j.test.resources.cache.dir=${GITHUB_WORKSPACE}/cache -DexcludedGroups="long-running-tests, large-resources, distributed-systems"  -DskipTestResourceEnforcement=true -Ptestresources  -Pintegration-tests  -Pnd4j-tests-cuda   clean test --fail-never
          mvn -Dorg.nd4j.strumpf.resource.dirs=${GITHUB_WORKSPACE}/resources -Dorg.nd4j.test.resources.cache.dir=${GITHUB_WORKSPACE}/cache -Dgroups="long-running-tests, large-resources, distributed-systems" -Ptestresources -Pnd4j-tests-cuda  -Dtest.offheap.size=14g -Dtest.heap.size=6g  -Dsurefire.parallel.forcedTimeout=200 -Dsurefire.parallel.timeout=200  -Dsurefire.timeout=200 -Dsurefire.exitTimeout=200 test --fail-never -rf :nd4j
          echo "Running tests for cuda 11.2"
          ${GITHUB_WORKSPACE}/change-cuda-versions.sh 11.2
          export PATH="/opt/protobuf/bin:/usr/local/cuda-11.2/bin:$PATH"
          mvn -Dorg.nd4j.strumpf.resource.dirs=${GITHUB_WORKSPACE}/resources -Dorg.nd4j.test.resources.cache.dir=${GITHUB_WORKSPACE}/cache -DexcludedGroups="long-running-tests, large-resources, distributed-systems"  -DskipTestResourceEnforcement=true -Ptestresources  -Pintegration-tests  -Pnd4j-tests-cuda   clean test --fail-never
          mvn -Dorg.nd4j.strumpf.resource.dirs=${GITHUB_WORKSPACE}/resources -Dorg.nd4j.test.resources.cache.dir=${GITHUB_WORKSPACE}/cache -Dgroups="long-running-tests, large-resources, distributed-systems" -Ptestresources -Pnd4j-tests-cuda  -Dtest.offheap.size=14g -Dtest.heap.size=6g  -Dsurefire.parallel.forcedTimeout=200 -Dsurefire.parallel.timeout=200  -Dsurefire.timeout=200 -Dsurefire.exitTimeout=200 test --fail-never -rf :nd4j