Skip to content

Optimize Decoder Pipeline Model Execution #2450

Optimize Decoder Pipeline Model Execution

Optimize Decoder Pipeline Model Execution #2450

name: "MacOS CPU ARM64 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"
jobs:
mac-cpu-arm64-build:
runs-on: macos-latest # arm64
steps:
- name: Checkout OnnxRuntime GenAI repo
uses: actions/checkout@v4
with:
submodules: true
- name: Get the Latest OnnxRuntime Nightly Version
run: |
ORT_NIGHTLY_VERSION=$(curl -s "${{ env.ORT_NIGHTLY_REST_API }}" | jq -r '.value[0].versions[0].normalizedVersion')
echo "$ORT_NIGHTLY_VERSION"
echo "ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" >> $GITHUB_ENV
- name: Download OnnxRuntime Nightly
run: |
nuget install ${{ env.ORT_PACKAGE_NAME }} -version ${{ env.ORT_NIGHTLY_VERSION }} -x
- name: Extract OnnxRuntime library and header files
run: |
mkdir -p ort/lib
mv ${{ env.ORT_PACKAGE_NAME }}/build/native/include ort/
mv ${{ env.ORT_PACKAGE_NAME }}/runtimes/osx-arm64/native/* ort/lib/
- name: Configure CMake
run: |
cmake --preset macos_arm64_cpu_release
- name: Build with CMake
run: |
cmake --build --preset macos_arm64_cpu_release --parallel
continue-on-error: false
- name: Install the python wheel and test dependencies
run: |
python3 -m venv genai-macos-venv
source genai-macos-venv/bin/activate
python3 -m pip install -r test/python/requirements.txt
python3 -m pip install -r test/python/requirements-macos.txt
python3 -m pip install build/cpu/osx-arm64/wheel/onnxruntime_genai*.whl --no-deps
- name: Remove the ort lib and header files
run: |
rm -rf ort
- name: Verify Build Artifacts
if: always()
continue-on-error: true
run: |
ls -l ${{ github.workspace }}/build/cpu/osx-arm64
# This will also download all the test models to the test/test_models directory
# These models are used by the python tests as well as C#, C++ and others.
- name: Run the python tests
run: |
source genai-macos-venv/bin/activate
export ORTGENAI_LOG_ORT_LIB=1
python3 test/python/test_onnxruntime_genai.py --cwd test/python --test_models test/test_models
- name: Build the C# API and Run the C# Tests
run: |
export ORTGENAI_LOG_ORT_LIB=1
cd test/csharp
dotnet test /p:Configuration=Release /p:NativeBuildOutputDir="../../build/cpu/osx-arm64"
- name: Run tests
run: |
set -e -x
export ORTGENAI_LOG_ORT_LIB=1
export DYLD_LIBRARY_PATH=$DYLD_LIBRARY_PATH:$GITHUB_WORKSPACE/build/cpu/osx-arm64
./build/cpu/osx-arm64/test/unit_tests