wehub-resource-sync
323 lines
11 KiB
YAML
323 lines
11 KiB
YAML
name: PR-CI-Inference
|
|
|
|
on:
|
|
workflow_call:
|
|
inputs:
|
|
docker_inference_image:
|
|
type: string
|
|
required: true
|
|
clone-can-skip:
|
|
type: string
|
|
required: false
|
|
default: "false"
|
|
|
|
env:
|
|
PR_ID: ${{ github.event.pull_request.number }}
|
|
COMMIT_ID: ${{ github.event.pull_request.head.sha }}
|
|
ci_scripts: /paddle/ci
|
|
work_dir: /paddle
|
|
PADDLE_ROOT: /paddle
|
|
BRANCH: ${{ github.event.pull_request.base.ref }}
|
|
CFS_DIR: /home/data/cfs
|
|
no_proxy: "bcebos.com,apiin.im.baidu.com,gitee.com,aliyun.com,.baidu.com,.tuna.tsinghua.edu.cn"
|
|
docker_image: ${{ inputs.docker_inference_image }}
|
|
|
|
defaults:
|
|
run:
|
|
shell: bash
|
|
|
|
jobs:
|
|
check-bypass:
|
|
name: Check bypass
|
|
uses: ./.github/workflows/check-bypass.yml
|
|
with:
|
|
workflow-name: "inference"
|
|
secrets:
|
|
github-token: ${{ secrets.GITHUB_TOKEN }}
|
|
|
|
build:
|
|
name: Build
|
|
needs: [check-bypass]
|
|
if: ${{ needs.check-bypass.outputs.can-skip != 'true' && inputs.clone-can-skip != 'true' }}
|
|
env:
|
|
TASK: paddle-CI-${{ github.event.pull_request.number }}-inference_build
|
|
runs-on:
|
|
group: GZ_BD-CPU
|
|
timeout-minutes: 120
|
|
|
|
steps:
|
|
- name: Check docker image and run container
|
|
env:
|
|
FLAGS_fraction_of_gpu_memory_to_use: "0.15"
|
|
CTEST_OUTPUT_ON_FAILURE: 1
|
|
CTEST_PARALLEL_LEVEL: 4
|
|
WITH_GPU: "ON"
|
|
WITH_AVX: "ON"
|
|
WITH_TESTING: "OFF"
|
|
COVERALLS_UPLOAD: "ON"
|
|
PADDLE_VERSION: 0.0.0
|
|
CUDA_VISIBLE_DEVICES: "0,1"
|
|
GIT_PR_ID: ${{ github.event.pull_request.number }}
|
|
INFERENCE_DEMO_INSTALL_DIR: /root/.cache/inference_demo
|
|
WITH_PYTHON: "OFF"
|
|
PY_VERSION: "3.10"
|
|
WITH_ONNXRUNTIME: "ON"
|
|
WITH_UNITY_BUILD: "ON"
|
|
WITH_NVCC_LAZY: "OFF"
|
|
PYTHON_ABI: cp310-cp310
|
|
CACHE_DIR: /root/.cache/inference
|
|
CCACHE_DIR: "/home/data/shared/.ccache/l1" # L1 cache on machine shared dir
|
|
CCACHE_SECONDARY_STORAGE: "file:///home/data/cfs/.ccache/l2" # L2 cache on cfs
|
|
CCACHE_MAXSIZE: 50G
|
|
CCACHE_LIMIT_MULTIPLE: "0.8"
|
|
CUDA_ARCH_NAME: Pascal
|
|
PADDLE_CUDA_INSTALL_REQUIREMENTS: "ON"
|
|
run: |
|
|
container_name=${TASK}-$(date +%Y%m%d-%H%M%S)
|
|
echo "container_name=${container_name}" >> ${{ github.env }}
|
|
docker run -d -t --name ${container_name} \
|
|
-v "/home/data/cfs:/home/data/cfs" \
|
|
-v "/home/data/cfs/.cache:/root/.cache" \
|
|
-v "/home/data/cfs/.ccache:/root/.ccache" \
|
|
-v "/home/data/cfs/.ccache/inference:/root/.ccache/inference" \
|
|
-v "/dev/shm:/dev/shm" \
|
|
-v "${{ github.workspace }}/../../..:/action-runner" \
|
|
-e BRANCH \
|
|
-e PR_ID \
|
|
-e COMMIT_ID \
|
|
-e work_dir \
|
|
-e PADDLE_ROOT \
|
|
-e FLAGS_fraction_of_gpu_memory_to_use \
|
|
-e CTEST_OUTPUT_ON_FAILURE \
|
|
-e CTEST_PARALLEL_LEVEL \
|
|
-e WITH_GPU \
|
|
-e WITH_AVX \
|
|
-e CUDA_VISIBLE_DEVICES \
|
|
-e INFERENCE_DEMO_INSTALL_DIR \
|
|
-e WITH_PYTHON \
|
|
-e WITH_ONNXRUNTIME \
|
|
-e WITH_NVCC_LAZY \
|
|
-e PYTHON_ABI \
|
|
-e COVERALLS_UPLOAD \
|
|
-e GIT_PR_ID \
|
|
-e PADDLE_VERSION \
|
|
-e WITH_TESTING \
|
|
-e PY_VERSION \
|
|
-e WITH_UNITY_BUILD \
|
|
-e CACHE_DIR \
|
|
-e CCACHE_DIR \
|
|
-e CCACHE_MAXSIZE \
|
|
-e CCACHE_LIMIT_MULTIPLE \
|
|
-e ci_scripts \
|
|
-e CUDA_ARCH_NAME \
|
|
-e PADDLE_CUDA_INSTALL_REQUIREMENTS \
|
|
-e CFS_DIR \
|
|
-e no_proxy \
|
|
-w /paddle --network host ${docker_image} /bin/bash
|
|
|
|
- name: Download paddle.tar.gz and update test branch
|
|
timeout-minutes: 30
|
|
run: |
|
|
docker exec -t ${container_name} /bin/bash -c '
|
|
rm -rf * .[^.]*
|
|
set -e
|
|
echo "Downloading Paddle.tar.gz"
|
|
wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PR/Paddle-build/${PR_ID}/${COMMIT_ID}/Paddle.tar.gz --no-check-certificate
|
|
echo "Extracting Paddle.tar.gz"
|
|
tar -xf Paddle.tar.gz --strip-components=1
|
|
rm Paddle.tar.gz
|
|
git config --global user.name "PaddleCI"
|
|
git config --global user.email "paddle_ci@example.com"
|
|
git config --global --add safe.directory "*"
|
|
git remote -v
|
|
set +e
|
|
git remote add upstream https://github.com/PaddlePaddle/Paddle.git
|
|
set -e
|
|
source /action-runner/proxy
|
|
git config pull.rebase false
|
|
git checkout test
|
|
echo "Pull upstream develop"
|
|
bash ci/git_pull.sh $BRANCH
|
|
git submodule update
|
|
'
|
|
|
|
- name: Run build
|
|
run: |
|
|
docker exec -t ${container_name} /bin/bash -c '
|
|
source /action-runner/proxy
|
|
bash ${ci_scripts}/cmake-predownload.sh
|
|
bash ${ci_scripts}/inference_build.sh
|
|
EXCODE=$?
|
|
exit $EXCODE
|
|
'
|
|
|
|
- name: Upload build.tar.gz
|
|
env:
|
|
home_path: /action-runner
|
|
bos_file: /action-runner/bos_retry/BosClient.py
|
|
run: |
|
|
docker exec -t ${{ env.container_name }} /bin/bash -c '
|
|
echo "::group::Install bce-python-sdk"
|
|
source /action-runner/proxy
|
|
python -m pip install bce-python-sdk==0.8.74
|
|
echo "::endgroup::"
|
|
rm -rf `find . -name "*.o"`
|
|
export AK=paddle
|
|
export SK=paddle
|
|
if [ ! -f "${{ env.bos_file }}" ]; then
|
|
wget -q --tries=5 --no-proxy -O ${{ env.home_path }}/bos_retry.tar.gz https://xly-devops.bj.bcebos.com/home/bos_retry.tar.gz --no-check-certificate
|
|
mkdir ${{ env.home_path }}/bos_retry
|
|
tar xf ${{ env.home_path }}/bos_retry.tar.gz -C ${{ env.home_path }}/bos_retry
|
|
fi
|
|
source /action-runner/unproxy
|
|
cd /
|
|
tar --use-compress-program="pigz -1" -cpPf build.tar.gz paddle
|
|
echo "Uploading build.tar.gz to cfs"
|
|
mkdir -p ${CFS_DIR}/inference_bos/${PR_ID}/${COMMIT_ID}
|
|
cp build.tar.gz ${CFS_DIR}/inference_bos/${PR_ID}/${COMMIT_ID} && echo success
|
|
rm build.tar.gz
|
|
'
|
|
|
|
- name: Terminate and delete the container
|
|
if: always()
|
|
run: |
|
|
# docker exec -t ${container_name} /bin/bash -c 'rm -rf * .[^.]*'
|
|
docker stop ${container_name}
|
|
docker rm ${container_name}
|
|
|
|
test:
|
|
name: Test
|
|
needs: build
|
|
env:
|
|
TASK: paddle-CI-${{ github.event.pull_request.number }}-inference_test
|
|
runs-on:
|
|
group: Inference-P4
|
|
timeout-minutes: 120
|
|
|
|
steps:
|
|
- name: Check docker image and run container
|
|
env:
|
|
FLAGS_fraction_of_gpu_memory_to_use: "0.15"
|
|
CTEST_OUTPUT_ON_FAILURE: 1
|
|
CTEST_PARALLEL_LEVEL: 4
|
|
WITH_GPU: "ON"
|
|
WITH_AVX: "ON"
|
|
WITH_TESTING: "OFF"
|
|
COVERALLS_UPLOAD: "ON"
|
|
PADDLE_VERSION: 0.0.0
|
|
CUDA_VISIBLE_DEVICES: "0,1"
|
|
GIT_PR_ID: ${{ github.event.pull_request.number }}
|
|
INFERENCE_DEMO_INSTALL_DIR: /root/.cache/inference_demo
|
|
WITH_PYTHON: "OFF"
|
|
PY_VERSION: "3.10"
|
|
WITH_ONNXRUNTIME: "ON"
|
|
WITH_UNITY_BUILD: "ON"
|
|
WITH_NVCC_LAZY: "OFF"
|
|
PYTHON_ABI: cp310-cp310
|
|
CACHE_DIR: /root/.cache/inference
|
|
CCACHE_DIR: /root/.ccache/inference
|
|
CCACHE_MAXSIZE: 150G
|
|
CCACHE_LIMIT_MULTIPLE: "0.8"
|
|
CUDA_ARCH_NAME: Auto
|
|
PADDLE_CUDA_INSTALL_REQUIREMENTS: "ON"
|
|
GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
|
run: |
|
|
container_name=${TASK}-$(date +%Y%m%d-%H%M%S)
|
|
echo "container_name=${container_name}" >> ${{ github.env }}
|
|
docker pull $docker_image
|
|
docker run -d -t --gpus all --name ${container_name} \
|
|
-v "/home/data/cfs:/home/data/cfs" \
|
|
-v "/home/data/cfs/.cache:/root/.cache" \
|
|
-v "/home/data/shared:/home/data/shared" \
|
|
-v "/dev/shm:/dev/shm" \
|
|
-v "${{ github.workspace }}/../../..:/action-runner" \
|
|
-e BRANCH \
|
|
-e PR_ID \
|
|
-e COMMIT_ID \
|
|
-e work_dir \
|
|
-e PADDLE_ROOT \
|
|
-e FLAGS_fraction_of_gpu_memory_to_use \
|
|
-e CTEST_OUTPUT_ON_FAILURE \
|
|
-e CTEST_PARALLEL_LEVEL \
|
|
-e WITH_GPU \
|
|
-e WITH_AVX \
|
|
-e CUDA_VISIBLE_DEVICES \
|
|
-e INFERENCE_DEMO_INSTALL_DIR \
|
|
-e WITH_PYTHON \
|
|
-e WITH_ONNXRUNTIME \
|
|
-e WITH_NVCC_LAZY \
|
|
-e PYTHON_ABI \
|
|
-e COVERALLS_UPLOAD \
|
|
-e GIT_PR_ID \
|
|
-e PADDLE_VERSION \
|
|
-e WITH_TESTING \
|
|
-e PY_VERSION \
|
|
-e WITH_UNITY_BUILD \
|
|
-e CACHE_DIR \
|
|
-e CCACHE_DIR \
|
|
-e CCACHE_SECONDARY_STORAGE \
|
|
-e CCACHE_MAXSIZE \
|
|
-e CCACHE_LIMIT_MULTIPLE \
|
|
-e ci_scripts \
|
|
-e CUDA_ARCH_NAME \
|
|
-e PADDLE_CUDA_INSTALL_REQUIREMENTS \
|
|
-e GITHUB_API_TOKEN \
|
|
-e CFS_DIR \
|
|
-e no_proxy \
|
|
-w /paddle --network host ${docker_image} /bin/bash
|
|
|
|
- name: Download paddle.tar.gz
|
|
timeout-minutes: 30
|
|
run: |
|
|
docker exec -t ${{ env.container_name }} /bin/bash -c '
|
|
set -e
|
|
echo "Downloading build.tar.gz from cfs"
|
|
cp ${CFS_DIR}/inference_bos/${PR_ID}/${COMMIT_ID}/build.tar.gz .
|
|
echo "Extracting build.tar.gz"
|
|
tar --use-compress-program="pigz -1" -xpf build.tar.gz --strip-components=1
|
|
rm build.tar.gz
|
|
export PATH=$(pwd)/run_env:${PATH}
|
|
echo "export PATH=${PATH}" >> ~/.bashrc
|
|
echo "::group::Install dependencies"
|
|
source /action-runner/proxy
|
|
pip install -r "python/unittest_py/requirements.txt"
|
|
echo "::endgroup::"
|
|
'
|
|
|
|
- name: Test fluid library for inference
|
|
run: |
|
|
docker exec -t ${{ env.container_name }} /bin/bash -c '
|
|
source ~/.bashrc
|
|
source ${ci_scripts}/utils.sh
|
|
init
|
|
source /action-runner/proxy
|
|
test_fluid_lib
|
|
'
|
|
|
|
- name: Test go inference api
|
|
run: |
|
|
docker exec -t ${{ env.container_name }} /bin/bash -c '
|
|
source ~/.bashrc
|
|
source ${ci_scripts}/utils.sh
|
|
init
|
|
source /action-runner/proxy
|
|
test_go_inference_api
|
|
'
|
|
|
|
- name: Check approvals of unittest
|
|
run: |
|
|
docker exec -t ${{ env.container_name }} /bin/bash -c '
|
|
source ~/.bashrc
|
|
source ${ci_scripts}/utils.sh
|
|
init
|
|
source /action-runner/proxy
|
|
check_approvals_of_unittest 3
|
|
'
|
|
|
|
- name: Terminate and delete the container
|
|
if: always()
|
|
run: |
|
|
# docker exec -t ${container_name} /bin/bash -c 'rm -rf * .[^.]*'
|
|
docker stop ${container_name}
|
|
docker rm ${container_name}
|