Add graph capture and shaders to improve DirectML performance #1351
Workflow file for this run
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: "Linux CUDA x64 Build" | |
| on: [ workflow_dispatch, pull_request ] | |
| concurrency: | |
| group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }} | |
| cancel-in-progress: true | |
| env: | |
| ort_dir: "onnxruntime-linux-x64-gpu-1.17.3" | |
| ort_zip: "onnxruntime-linux-x64-gpu-1.17.3.tgz" | |
| ort_url: "https://github.com/microsoft/onnxruntime/releases/download/v1.17.3/onnxruntime-linux-x64-gpu-1.17.3.tgz" | |
| jobs: | |
| linux-cuda-x64-build: | |
| env : | |
| PYTHON_EXECUTABLE: "/opt/python/cp38-cp38/bin/python3.8" | |
| runs-on: [ "self-hosted", "1ES.Pool=onnxruntime-genai-Ubuntu2004-T4" ] | |
| steps: | |
| - name: Checkout OnnxRuntime GenAI repo | |
| uses: actions/checkout@v4 | |
| with: | |
| submodules: true | |
| # We are using the same manylinux repo as the one used in the packaging build | |
| - name: Checkout ManyLinux repo | |
| uses: actions/checkout@v4 | |
| with: | |
| repository: pypa/manylinux | |
| ref: 5eda9aded5462201e6310105728d33016e637ea7 | |
| clean: true | |
| path: manylinux | |
| submodules: true | |
| - name: Download OnnxRuntime | |
| run: | | |
| curl -L -o ${{ env.ort_zip }} ${{ env.ort_url }} | |
| - name: Unzip OnnxRuntime | |
| run: | | |
| tar -xzf ${{ env.ort_zip }} | |
| rm ${{ env.ort_zip }} | |
| - name: Rename OnnxRuntime to ort | |
| run: | | |
| mv ${{ env.ort_dir }} ort | |
| - name: Get Docker Image | |
| run: | | |
| set -e -x | |
| az login --identity --username 63b63039-6328-442f-954b-5a64d124e5b4 | |
| az acr login --name onnxruntimebuildcache --subscription 00c06639-6ee4-454e-8058-8d8b1703bd87 | |
| python3 tools/ci_build/get_docker_image.py --dockerfile tools/ci_build/github/linux/docker/manylinux/Dockerfile.manylinux2_28_cuda \ | |
| --context tools/ci_build/github/linux/docker/manylinux \ | |
| --docker-build-args "--build-arg BUILD_UID=$( id -u )" \ | |
| --container-registry onnxruntimebuildcache \ | |
| --manylinux-src manylinux \ | |
| --multiple_repos \ | |
| --repository onnxruntimecudabuildx64 | |
| - name: Config with Cmake in Docker | |
| run: | | |
| set -e -x | |
| docker run \ | |
| --gpus all \ | |
| --rm \ | |
| --volume $GITHUB_WORKSPACE:/ort_genai_src \ | |
| -w /ort_genai_src onnxruntimecudabuildx64 \ | |
| bash -c " \ | |
| /usr/bin/cmake --preset linux_gcc_cuda_release \ | |
| -DMANYLINUX=ON \ | |
| -DPYTHON_EXECUTABLE=${{ env.PYTHON_EXECUTABLE }} " | |
| - name: Build with Cmake in Docker | |
| run: | | |
| set -e -x | |
| docker run \ | |
| --gpus all \ | |
| --rm \ | |
| --volume $GITHUB_WORKSPACE:/ort_genai_src \ | |
| -w /ort_genai_src onnxruntimecudabuildx64 \ | |
| bash -c " \ | |
| /usr/bin/cmake --build --preset linux_gcc_cuda_release --parallel $( nproc )" | |
| - name: Get HuggingFace Token | |
| run: | | |
| az login --identity --username 63b63039-6328-442f-954b-5a64d124e5b4 | |
| HF_TOKEN=$(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: Install the onnxruntime-genai Python wheel and run python test | |
| run: | | |
| echo "Installing the onnxruntime-genai Python wheel and running the Python tests" | |
| docker run \ | |
| --gpus all \ | |
| --rm \ | |
| --volume $GITHUB_WORKSPACE:/ort_genai_src \ | |
| -e HF_TOKEN=$HF_TOKEN \ | |
| -w /ort_genai_src onnxruntimecudabuildx64 bash -c " \ | |
| ${{ env.PYTHON_EXECUTABLE }} -m pip install /ort_genai_src/build/cuda/wheel/onnxruntime_genai*manylinux*.whl --user && \ | |
| ${{ env.PYTHON_EXECUTABLE }} -m pip install -r test/python/requirements-gpu.txt --user && \ | |
| ${{ env.PYTHON_EXECUTABLE }} test/python/test_onnxruntime_genai.py --cwd test/python --test_models test/test_models" | |
| - name: Docker -- Run unit tests | |
| run: | | |
| echo "Running docker image onnxruntimecudabuildx64" | |
| docker run \ | |
| --gpus all \ | |
| --rm \ | |
| --volume $GITHUB_WORKSPACE:/ort_genai_src \ | |
| -w /ort_genai_src onnxruntimecudabuildx64 bash -c "/ort_genai_src/build/cuda/test/unit_tests" |