Skip to content

cpu top k and top p+k / test cases for cpu sampling #402

cpu top k and top p+k / test cases for cpu sampling

cpu top k and top p+k / test cases for cpu sampling #402

name: "Linux GPU 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-1.17.0"
ort_zip: "onnxruntime-linux-x64-gpu-1.17.0.tgz"
ort_url: "https://github.com/microsoft/onnxruntime/releases/download/v1.17.0/onnxruntime-linux-x64-gpu-1.17.0.tgz"
jobs:
job:
runs-on: [ "self-hosted", "1ES.Pool=onnxruntime-genai-Ubuntu2004-T4" ]
steps:
- name: Checkout OnnxRuntime GenAI repo
uses: actions/checkout@v2
- 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 ${{ env.ort_dir }} ort
- name: Git Submodule Update
run: |
git submodule update --init --recursive
- name: Download Docker Image
run: |
set -e -x
python3 tools/ci_build/get_docker_image.py --dockerfile tools/ci_build/github/linux/docker/inference/x64/default/gpu/Dockerfile \
--context tools/ci_build/github/linux/docker/inference/x64/default/gpu \
--docker-build-args "--build-arg BUILD_UID=$( id -u )" \
--container-registry onnxruntimebuildcache \
--repository onnxruntimegpubuild
- name: Print Docker Image Environment Variables
run: |
echo "Printing docker image environment variables"
docker run --rm onnxruntimegpubuild env
- name: Build with Cmake in Docker
run: |
echo "Running docker image onnxruntimegpubuild"
docker run \
--gpus all \
--rm \
--volume $GITHUB_WORKSPACE:/onnxruntime_src \
-w /onnxruntime_src onnxruntimegpubuild bash -c "echo $PATH && /usr/bin/cmake -DCMAKE_CUDA_ARCHITECTURES=86 --preset linux_gcc_cuda_release && /usr/bin/cmake --build --preset linux_gcc_cuda_release"