Skip to content

Add GroupQueryAttention on CPU in model builder (#420) #1

Add GroupQueryAttention on CPU in model builder (#420)

Add GroupQueryAttention on CPU in model builder (#420) #1

name: "Windows CUDA x64 Build"
on:
workflow_dispatch:
push:
branches:
- main
- rel-*
pull_request:
concurrency:
group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
cancel-in-progress: true
env:
AZCOPY_AUTO_LOGIN_TYPE: MSI
AZCOPY_MSI_CLIENT_ID: 63b63039-6328-442f-954b-5a64d124e5b4
cuda_dir: "${{ github.workspace }}\\cuda_sdk"
cuda_version: "11.8"
CUDA_PATH: ${{ github.workspace }}\\cuda_sdk\\v11.8
binaryDir: 'build/cuda'
ORT_NIGHTLY_REST_API: "https://feeds.dev.azure.com/aiinfra/PublicPackages/_apis/packaging/Feeds/ORT-Nightly/packages?packageNameQuery=Microsoft.ML.OnnxRuntime.Gpu.Windows&api-version=6.0-preview.1"
ORT_PACKAGE_NAME: "Microsoft.ML.OnnxRuntime.Gpu.Windows"
jobs:
windows-cuda-x64-build:
runs-on: [ "self-hosted", "1ES.Pool=onnxruntime-genai-Win2022-GPU-A10" ]
steps:
- name: Checkout OnnxRuntime GenAI repo
uses: actions/checkout@v4
with:
submodules: true
- uses: actions/setup-python@v5
with:
python-version: '3.11.x'
architecture: 'x64'
- name: Download cuda
run: |
azcopy.exe cp --recursive "https://lotusscus.blob.core.windows.net/models/cuda_sdk/v${{ env.cuda_version }}" ${{ env.cuda_dir}}
- uses: actions/setup-dotnet@v4
with:
dotnet-version: '6.0.x'
- name : Install jq and curl
run: |
choco install -y jq curl
- name: Get the Latest OnnxRuntime Nightly Version
shell: pwsh
run: |
$ORT_NIGHTLY_VERSION=$(curl -s "${{ env.ORT_NIGHTLY_REST_API }}" | jq -r '.value[0].versions[0].normalizedVersion')
echo "$ORT_NIGHTLY_VERSION"
"ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" | Out-File -FilePath $env:GITHUB_ENV -Append
- name: Download OnnxRuntime Nightly
run: |
nuget install ${{ env.ORT_PACKAGE_NAME }} -version ${{ env.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 ${{ env.ORT_PACKAGE_NAME }}/runtimes/win-x64/native/* ort/lib/
- name: Configure CMake
run: |
cmake --preset windows_x64_cuda_release -T cuda=${{ env.cuda_dir }}\\v${{ env.cuda_version }} -DTEST_PHI2=False
- name: Build with CMake
run: |
cmake --build --preset windows_x64_cuda_release --parallel
- name: Add CUDA to PATH
run: |
echo "${{ env.cuda_dir }}\\v${{ env.cuda_version }}\\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
- name: Build the C# API and Run the C# Tests
run: |
cd test\csharp
dotnet test /p:Configuration=release /p:NativeBuildOutputDir="$env:GITHUB_WORKSPACE\$env:binaryDir\Release"
- name: Install the Python Wheel and Test Dependencies
run: |
python -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl"))
python -m pip install -r test\python\requirements-nightly-cpu.txt
- name: Get HuggingFace Token
run: |
az login --identity --username 63b63039-6328-442f-954b-5a64d124e5b4
$HF_TOKEN = (az keyvault secret show --vault-name anubissvcsecret --name ANUBIS-HUGGINGFACE-TOKEN --query value)
Write-Output "::add-mask::$HF_TOKEN"
Add-Content -Path $env:GITHUB_ENV -Value "HF_TOKEN=$HF_TOKEN"
- name: Run the Python Tests
run: |
python test/python/test_onnxruntime_genai.py --cwd "test\python" --test_models "test\test_models"
- name: Verify Build Artifacts
if: always()
continue-on-error: true
run: |
Get-ChildItem -Path $env:GITHUB_WORKSPACE\$env:binaryDir -Recurse
- 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"
.\build\cuda\test\Release\unit_tests.exe