win-cuda-x64-build.yml

name: "Windows CUDA x64 Build"
on:
  workflow_dispatch:
  push:
    branches:
      - main
      - rel-*
  pull_request:

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

env:
  AZCOPY_AUTO_LOGIN_TYPE: MSI
  AZCOPY_MSI_CLIENT_ID: d712a4c7-a0cf-4e87-af75-31510eba0a8e
  cuda_dir: "${{ github.workspace }}\\cuda_sdk"
  cuda_version: "12.8"
  CUDA_PATH: ${{ github.workspace }}\\cuda_sdk\\v12.8
  binaryDir: 'build/cuda/win-x64'
  ORT_NIGHTLY_REST_API: "https://feeds.dev.azure.com/aiinfra/PublicPackages/_apis/packaging/Feeds/ORT-Nightly/packages?packageNameQuery=Microsoft.ML.OnnxRuntime.Gpu.Windows&api-version=6.0-preview.1"
  ORT_PACKAGE_NAME: "Microsoft.ML.OnnxRuntime.Gpu.Windows"

jobs:
  windows-cuda-x64-build:
    runs-on: ["self-hosted", "1ES.Pool=onnxruntime-genai-Win2022-GPU-A10"]
    steps:
      - name: Checkout OnnxRuntime GenAI repo
        uses: actions/checkout@v5
        with:
          submodules: true

      - uses: actions/setup-python@v6
        with:
          python-version: '3.11.x'
          architecture: 'x64'

      - name: Setup VCPKG
        uses: microsoft/onnxruntime-github-actions/setup-build-tools@v0.0.8
        with:
          vcpkg-version: '2025.03.19'
          vcpkg-hash: '17e96169cd3f266c4716fcdc1bb728e6a64f103941ece463a2834d50694eba4fb48f30135503fd466402afa139abc847ef630733c442595d1c34979f261b0114'
          cmake-version: '3.31.6'
          cmake-hash: '0f1584e8666cf4a65ec514bd02afe281caabf1d45d2c963f3151c41484f457386aa03273ab25776a670be02725354ce0b46f3a5121857416da37366342a833a0'
          add-cmake-to-path: 'true'
          disable-terrapin: 'false'

      - name: Download cuda
        run: |
          azcopy.exe cp --recursive "https://lotusscus.blob.core.windows.net/models/cuda_sdk/v${{ env.cuda_version }}" ${{ env.cuda_dir}}

      - uses: actions/setup-dotnet@v5
        with:
          dotnet-version: '8.0.x'

      - name: Download OnnxRuntime Nightly
        shell: pwsh
        run: |
          $resp = Invoke-RestMethod "${{ env.ORT_NIGHTLY_REST_API }}"
          $ORT_NIGHTLY_VERSION = "1.26.0"
          Write-Host "$ORT_NIGHTLY_VERSION"
          "ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" | Out-File -FilePath $env:GITHUB_ENV -Append
          nuget install ${{ env.ORT_PACKAGE_NAME }} -version $ORT_NIGHTLY_VERSION -ExcludeVersion -NonInteractive

      - run: Get-ChildItem  ${{ env.ORT_PACKAGE_NAME }} -Recurse
        continue-on-error: true

      - name: Extract OnnxRuntime library and header files
        run: |
          mkdir ort/lib
          move ${{ env.ORT_PACKAGE_NAME }}/buildTransitive/native/include ort/
          move ${{ env.ORT_PACKAGE_NAME }}/runtimes/win-x64/native/* ort/lib/

      - name: Install Rust Toolchain
        run: |
          $exePath = "$env:TEMP\rustup-init.exe"
          (New-Object Net.WebClient).DownloadFile('https://static.rust-lang.org/rustup/dist/x86_64-pc-windows-msvc/rustup-init.exe', $exePath)
          & $exePath -y --default-toolchain=1.86.0
          Add-Content $env:GITHUB_PATH "$env:USERPROFILE\.cargo\bin"

      # Restore sccache from a previous run so the C/C++ compile step gets cache hits.
      # restore-keys allow a partial match from any prior run on this branch or main.
      - name: Restore sccache
        id: restore-sccache
        uses: actions/cache/restore@v4
        with:
          path: ${{ github.workspace }}\sccache
          key: sccache-win-cuda-${{ hashFiles('**/CMakePresets.json', '**/CMakeLists.txt') }}
          restore-keys: |
            sccache-win-cuda-

      - name: Install sccache
        run: |
          $version = "0.9.0"
          $url = "https://github.com/mozilla/sccache/releases/download/v${version}/sccache-v${version}-x86_64-pc-windows-msvc.tar.gz"
          $dest = "$env:TEMP\sccache.tar.gz"
          Invoke-WebRequest -Uri $url -OutFile $dest
          tar -xzf $dest -C "$env:TEMP"
          $binDir = "$env:USERPROFILE\sccache_bin"
          New-Item -ItemType Directory -Force -Path $binDir | Out-Null
          Copy-Item "$env:TEMP\sccache-v${version}-x86_64-pc-windows-msvc\sccache.exe" "$binDir\sccache.exe"
          echo $binDir | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
          & "$binDir\sccache.exe" --version

      - name: Configure CMake
        run: |
          $env:SCCACHE_DIR = "${{ github.workspace }}\sccache"
          # Limit CUDA archs to SM 86 (matches the A10 runner GPU) to keep CI build
          # time and memory usage manageable. The full SM list in CMakePresets.json
          # is intended for release packaging, not CI validation.
          cmake --preset windows_x64_cuda_release -T cuda=${{ env.cuda_dir }}\\v${{ env.cuda_version }} -DTEST_PHI2=True -DTEST_QWEN_2_5=True -DENABLE_JAVA=ON -DUSE_GUIDANCE=ON `
            -DCMAKE_CUDA_ARCHITECTURES=86-real `
            -DCMAKE_C_COMPILER_LAUNCHER=sccache `
            -DCMAKE_CXX_COMPILER_LAUNCHER=sccache

      - name: Build with CMake
        run: |
          $env:SCCACHE_DIR = "${{ github.workspace }}\sccache"
          cmake --build --preset windows_x64_cuda_release --parallel
          cmake --build --preset windows_x64_cuda_release --target PyPackageBuild

      - name: Show sccache stats
        run: sccache --show-stats

      # Save sccache inline (before the test steps) so the cache is written even if
      # a later step kills the runner. Skip if we had a full key hit (nothing new to store).
      - name: Save sccache
        if: steps.restore-sccache.outputs.cache-hit != 'true'
        uses: actions/cache/save@v4
        with:
          path: ${{ github.workspace }}\sccache
          key: sccache-win-cuda-${{ hashFiles('**/CMakePresets.json', '**/CMakeLists.txt') }}

      - name: Add CUDA to PATH
        run: |
          echo "${{ env.cuda_dir }}\\v${{ env.cuda_version }}\\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append

      - name: Build the C Examples
        run: |
          cmake -S examples\c -B examples\c\build `
            -DMODEL_QA=ON -DMODEL_CHAT=ON -DMODEL_MM=ON -DWHISPER=ON -DNEMOTRON_SPEECH=ON `
            "-DORT_INCLUDE_DIR=$env:GITHUB_WORKSPACE\ort\include" `
            "-DORT_LIB_DIR=$env:GITHUB_WORKSPACE\ort\lib" `
            "-DOGA_INCLUDE_DIR=$env:GITHUB_WORKSPACE\src" `
            "-DOGA_LIB_DIR=$env:GITHUB_WORKSPACE\$env:binaryDir\Release"
          cmake --build examples\c\build --config Release

      - name: Install the Python Wheel and Test Dependencies
        run: |
          python -m pip install -r test\python\requirements.txt
          python -m pip install -r test\python\cuda\torch\requirements.txt
          python -m pip install -r test\python\cuda\ort\requirements.txt
          python -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl")) --no-deps

      - name: Run the Java tests
        run: |
          ctest --test-dir $env:binaryDir\src\java --build-config Release --verbose --timeout 10800

      - name: Run the Python Tests
        run: |
          python test/python/test_onnxruntime_genai.py --cwd "test\python" --test_models "test\models" --e2e

      - name: Verify Build Artifacts
        if: always()
        continue-on-error: true
        run: |2

          Get-ChildItem -Path $env:GITHUB_WORKSPACE\$env:binaryDir -Recurse

      - name: Build the C# API and Run the C# Tests
        run: |
          $env:PATH = "${{ env.cuda_dir }}\\v${{ env.cuda_version }}\\bin;" + $env:PATH
          cd test\csharp
          dotnet test /p:Configuration=release /p:NativeBuildOutputDir="$env:GITHUB_WORKSPACE\$env:binaryDir\Release" /p:OrtLibDir="$env:GITHUB_WORKSPACE\ort\lib"

      - name: Build the C# Examples
        run: |
          cd examples\csharp\ModelChat
          dotnet build -c Release
          cd ..\ModelMM
          dotnet build -c Release

      - name: Test the C# LLM Example with Tool Calling
        run: |
          python test\python\special_tokens.py -p test\models\qwen-2.5-0.5b\int4\cpu\tokenizer.json -s "<tool_call>" -e "</tool_call>"
          .\examples\csharp\ModelChat\bin\Release\net8.0\ModelChat.exe -m test\models\qwen-2.5-0.5b\int4\cpu\ -e cpu --response_format lark_grammar --tools_file test\models\tool-definitions\weather.json --tool_call_start "<tool_call>" --tool_call_end "</tool_call>" --user_prompt "What is the weather in Redmond, WA?" --tool_output --non_interactive --verbose

      - name: Prepend CUDA to PATH and Run tests
        run: |-
          $env:PATH = "${{ env.cuda_dir }}\\v${{ env.cuda_version }}\\bin;" + $env:PATH
          echo "Current PATH variable is: $env:PATH"
          copy $env:GITHUB_WORKSPACE\ort\lib\* .\$env:binaryDir\Release
          & .\$env:binaryDir\Release\unit_tests.exe