Skip to content

Optimize Decoder Pipeline Model Execution #2940

Optimize Decoder Pipeline Model Execution

Optimize Decoder Pipeline Model Execution #2940

name: "Linux CPU 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:
ORT_NIGHTLY_REST_API: "https://feeds.dev.azure.com/aiinfra/PublicPackages/_apis/packaging/Feeds/ORT-Nightly/packages?packageNameQuery=Microsoft.ML.OnnxRuntime&api-version=6.0-preview.1"
ORT_PACKAGE_NAME: "Microsoft.ML.OnnxRuntime"
ORT_NIGHTLY_SOURCE: "https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/nuget/v3/index.json"
DOTNET_INSTALL_DIR: "${{ github.workspace }}/dotnet"
jobs:
linux_cpu_x64:
runs-on: [ "self-hosted", "1ES.Pool=onnxruntime-genai-Ubuntu2204-AMD-CPU" ]
steps:
- name: Checkout OnnxRuntime GenAI repo
uses: actions/checkout@v4
with:
submodules: true
- uses: actions/setup-dotnet@v4
with:
dotnet-version: '8.0.x'
- name: Get the Latest OnnxRuntime Nightly Version
shell: pwsh
run: |
$resp = Invoke-RestMethod "${{ env.ORT_NIGHTLY_REST_API }}"
$ORT_NIGHTLY_VERSION = $resp.value[0].versions[0].normalizedVersion
Write-Host "$ORT_NIGHTLY_VERSION"
"ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" | Out-File -FilePath $env:GITHUB_ENV -Append
- name: Download OnnxRuntime Nightly
run: |
dotnet new console
dotnet add package ${{ env.ORT_PACKAGE_NAME }} --version ${{ env.ORT_NIGHTLY_VERSION }} --source ${{ env.ORT_NIGHTLY_SOURCE }} --package-directory .
dotnet build
continue-on-error: true
- name: list files
shell: bash
run: |
ls -l
ls -R ${{ env.ORT_PACKAGE_NAME }}
continue-on-error: true
# TODO: Find out why do we need to to have libonnxruntime.so.$ort_version
- name: Extract OnnxRuntime library and header files
run: |
set -e -x
mkdir -p ort/lib
mv microsoft.ml.onnxruntime/${{ env.ORT_NIGHTLY_VERSION }}/build/native/include ort/
mv microsoft.ml.onnxruntime/${{ env.ORT_NIGHTLY_VERSION }}/runtimes/linux-x64/native/* ort/lib/
cp ort/lib/libonnxruntime.so ort/lib/libonnxruntime.so.1
- name: Build with CMake and GCC
run: |
set -e -x
rm -rf build
cmake --preset linux_gcc_cpu_release
cmake --build --preset linux_gcc_cpu_release
- name: Install the python wheel and test dependencies
run: |
python3 -m pip install -r test/python/requirements.txt --user
python3 -m pip install -r test/python/requirements-cpu.txt --user
python3 -m pip install build/cpu/wheel/onnxruntime_genai*.whl --user --no-deps
- name: Get HuggingFace Token
run: |
az login --identity --username 63b63039-6328-442f-954b-5a64d124e5b4
HF_TOKEN=$(az keyvault secret show --vault-name anubissvcsecret --name ANUBIS-HUGGINGFACE-TOKEN --query value)
echo "::add-mask::$HF_TOKEN"
echo "HF_TOKEN=$HF_TOKEN" >> $GITHUB_ENV
- name: Verify Build Artifacts
if: always()
continue-on-error: true
run: |
ls -l ${{ github.workspace }}/build/cpu
# This will also download all the test models to the test/test_models directory
# These models are used by the python tests as well as C#, C++ and others.
- name: Run the python tests
run: |
export ORTGENAI_LOG_ORT_LIB=1
python3 test/python/test_onnxruntime_genai.py --cwd test/python --test_models test/test_models
- name: Build the C# API and Run the C# Tests
run: |
export ORTGENAI_LOG_ORT_LIB=1
cd test/csharp
dotnet test /p:Configuration=Release /p:NativeBuildOutputDir="../../build/cpu/" /p:OrtLibDir="../../ort/lib/"
- name: Run tests
run: |
set -e -x
export ORTGENAI_LOG_ORT_LIB=1
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$GITHUB_WORKSPACE/ort/lib
./build/cpu/test/unit_tests