onnxruntime/.github/workflows/linux_tensorrt_ci.yml at 9f5fc1f44a588350d56bd9c426ab9f4e5bf237f0 · microsoft/onnxruntime · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
name: Linux TensorRT CI

on:
  push:
    branches: [main, 'rel-*']
  pull_request:
    branches: [main, 'rel-*']
  workflow_dispatch:

concurrency:
  group: ${{ github.workflow }}-${{ github.event_name == 'pull_request' && github.ref || github.sha }}
  cancel-in-progress: true

permissions:
  contents: read
  packages: write
  attestations: write
  id-token: write

jobs:
  build-linux-TensorRT-x64-release:
    name: Build Linux TensorRT x64 Release
    # This job runs on a CPU node using the reusable build workflow
    uses: ./.github/workflows/reusable_linux_build.yml
    with:
      pool_name: "onnxruntime-github-Ubuntu2204-AMD-CPU" # Build pool
      build_config: Release
      architecture: x64
      dockerfile_path: tools/ci_build/github/linux/docker/Dockerfile.manylinux2_28_cuda
      docker_build_args: '--build-arg BASEIMAGE=onnxruntimebuildcache.azurecr.io/internal/azureml/onnxruntime/build/cuda12_x64_almalinux8_gcc14:20251017.1 --build-arg TRT_VERSION=10.14.1.48-1.cuda12.9 --network=host'
      docker_image_repo: onnxruntimetensorrt86gpubuild
      extra_build_flags: '--use_binskim_compliant_compile_flags --build_wheel --cuda_version=12.8 --cuda_home=/usr/local/cuda-12.8 --cudnn_home=/usr/local/cuda-12.8 --use_tensorrt --tensorrt_home /usr  --build_java --parallel --nvcc_threads 4 --flash_nvcc_threads 4 --cmake_extra_defines onnxruntime_QUICK_BUILD=ON --cmake_extra_defines CMAKE_CUDA_ARCHITECTURES=86 onnxruntime_BUILD_UNIT_TESTS=ON onnxruntime_ENABLE_CUDA_EP_INTERNAL_TESTS=ON'
      python_path_prefix: 'PATH=/opt/python/cp310-cp310/bin:$PATH'
      run_tests: false            # <<< Do not run tests in this job
      upload_build_output: true   # <<< Upload the build/Release directory
      execution_providers: 'cuda tensorrt'
      job_identifier: build-linux-TensorRT-x64-release
    secrets:
      GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} # Pass token for reusable workflow needs (e.g., docker build action)

  build-linux-TensorRT-CUDA-Minimal-x64-release:
    name: Build Linux TensorRT CUDA Minimal x64 Release
    # Build-only job for CUDA minimal build (no tests, no unit tests)
    uses: ./.github/workflows/reusable_linux_build.yml
    with:
      pool_name: "onnxruntime-github-Ubuntu2204-AMD-CPU"
      build_config: Release
      architecture: x64
      dockerfile_path: tools/ci_build/github/linux/docker/Dockerfile.manylinux2_28_cuda
      docker_build_args: '--build-arg BASEIMAGE=onnxruntimebuildcache.azurecr.io/internal/azureml/onnxruntime/build/cuda12_x64_almalinux8_gcc14:20251017.1 --build-arg TRT_VERSION=10.14.1.48-1.cuda12.9 --network=host'
      docker_image_repo: onnxruntimetensorrt86gpubuild
      extra_build_flags: '--use_binskim_compliant_compile_flags --build_wheel --cuda_version=12.8 --cuda_home=/usr/local/cuda-12.8 --cudnn_home=/usr/local/cuda-12.8 --use_tensorrt --tensorrt_home /usr --build_java --enable_cuda_minimal_build --disable_generation_ops --cmake_extra_defines CMAKE_CUDA_ARCHITECTURES=86 onnxruntime_BUILD_UNIT_TESTS=OFF'
      python_path_prefix: 'PATH=/opt/python/cp310-cp310/bin:$PATH'
      run_tests: false
      upload_build_output: false
      execution_providers: 'cuda tensorrt'
      job_identifier: build-linux-TensorRT-CUDA-Minimal-x64-release
    secrets:
      GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

  test-linux-TensorRT-x64-release:
    name: Test Linux TensorRT x64 Release
    needs: build-linux-TensorRT-x64-release
    runs-on:
      - self-hosted
      - "1ES.Pool=onnxruntime-github-linux-a10"
      - "1ES.ImageOverride=onnxruntime-ubuntu2204-CUDA-A10-Test"
      - "JobId=test-linux-TensorRT-x64-release-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}"
    timeout-minutes: 210
    permissions:
      contents: read
      packages: read
    steps:
      - name: Checkout code
        uses: actions/checkout@d23441a48e516b6c34aea4fa41551a30e30af803 # v6.1.0

      # --- Build the Docker image needed for testing ---
      - name: Build Docker Image for Testing
        uses: microsoft/onnxruntime-github-actions/build-docker-image@8bad63a3c05d448311dfa8e5f531171c97471aa1 # v0.0.12
        id: build_docker_image_step
        with:
          dockerfile: ${{ github.workspace }}/tools/ci_build/github/linux/docker/Dockerfile.manylinux2_28_cuda
          image-name: ghcr.io/microsoft/onnxruntime/onnxruntimetensorrt86gpubuild
          build-args: '--build-arg BASEIMAGE=onnxruntimebuildcache.azurecr.io/internal/azureml/onnxruntime/build/cuda12_x64_almalinux8_gcc14:20251017.1 --build-arg TRT_VERSION=10.14.1.48-1.cuda12.9 --network=host'
          push: true
          azure-container-registry-name: onnxruntimebuildcache
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} # Pass token to action

      # --- Download Build Artifact to Runner Temp Directory ---
      - name: Download Build Artifact
        uses: actions/download-artifact@37930b1c2abaa49bbe596cd826c3c89aef350131 # v7.0.0
        with:
          name: build-output-x64-Release # Must match the upload name
          path: ${{ runner.temp }}/Release # Download contents into temp dir structure

      # --- Restore Permissions in the Temp Directory ---
      - name: Restore Executable Permissions
        if: success() # Only run if download succeeded
        working-directory: ${{ runner.temp }}/Release
        run: |
          if [ -f perms.txt ]; then
            echo "Restoring executable permissions in ${{ runner.temp }}/Release ..."
            while IFS= read -r file; do
              # Check relative path existence within the current directory
              if [ -f "$file" ]; then
                chmod +x "$file"
              else
                echo "Warning: File '$file' listed in perms.txt not found."
              fi
            done < perms.txt
            echo "Permissions restored."
          else
            echo "Warning: perms.txt not found in artifact."
          fi

      # --- Run Tests using the downloaded build ---
      # The run-build-script-in-docker action mounts ${{ runner.temp }} to /onnxruntime_src/build
      # So build.py --build_dir build/Release inside the container correctly finds the artifacts.
      - name: Test ONNX Runtime
        id: test_step
        uses: microsoft/onnxruntime-github-actions/run-build-script-in-docker@8bad63a3c05d448311dfa8e5f531171c97471aa1 # v0.0.12
        with:
          docker_image: ${{ steps.build_docker_image_step.outputs.full-image-name }}
          build_config: Release
          mode: 'test' # Set mode to test
          execution_providers: 'cuda tensorrt'
          extra_build_flags: '--use_binskim_compliant_compile_flags --build_wheel --cuda_version=12.8 --cuda_home=/usr/local/cuda-12.8 --cudnn_home=/usr/local/cuda-12.8 --use_tensorrt --tensorrt_home /usr  --build_java --cmake_extra_defines CMAKE_CUDA_ARCHITECTURES=86 onnxruntime_BUILD_UNIT_TESTS=ON onnxruntime_ENABLE_CUDA_EP_INTERNAL_TESTS=ON'
          python_path_prefix: 'PATH=/opt/python/cp310-cp310/bin:$PATH'