wehub-resource-sync
198 lines
7.3 KiB
YAML
198 lines
7.3 KiB
YAML
name: Linux-NPU
|
|
|
|
on:
|
|
workflow_call:
|
|
inputs:
|
|
can-skip:
|
|
type: string
|
|
required: false
|
|
default: "false"
|
|
docker_npu_image:
|
|
type: string
|
|
required: true
|
|
|
|
env:
|
|
docker_image: ${{ inputs.docker_npu_image }}
|
|
PR_ID: ${{ github.event.pull_request.number }}
|
|
COMMIT_ID: ${{ github.event.pull_request.head.sha }}
|
|
ci_scripts: /paddle/ci
|
|
ci_scripts_runner: ${{ github.workspace }}/ci
|
|
work_dir: /paddle
|
|
PADDLE_ROOT: /paddle
|
|
TASK: paddle-CI-${{ github.event.pull_request.number }}-npu
|
|
BRANCH: ${{ github.event.pull_request.base.ref }}
|
|
CI_name: npu
|
|
|
|
jobs:
|
|
check-bypass:
|
|
name: Check bypass
|
|
uses: ./.github/workflows/check-bypass.yml
|
|
with:
|
|
workflow-name: "npu"
|
|
secrets:
|
|
github-token: ${{ secrets.GITHUB_TOKEN }}
|
|
|
|
test:
|
|
name: Test
|
|
needs: [check-bypass]
|
|
if: ${{ needs.check-bypass.outputs.can-skip != 'true' && inputs.can-skip != 'true' }}
|
|
runs-on:
|
|
group: NPU
|
|
timeout-minutes: 120
|
|
|
|
steps:
|
|
- name: Download paddle.tar.gz and update test branch
|
|
timeout-minutes: 30
|
|
run: |
|
|
docker run -i --rm \
|
|
-v ${{ github.workspace }}:/paddle \
|
|
-e PR_ID \
|
|
-e COMMIT_ID \
|
|
-e BRANCH \
|
|
-w /paddle $docker_image /bin/bash -c 'rm -rf * .[^.]*'
|
|
source ~/.bashrc
|
|
set -e
|
|
echo "Downloading Paddle.tar.gz"
|
|
wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PR/Paddle/${PR_ID}/${COMMIT_ID}/Paddle.tar.gz --no-check-certificate
|
|
echo "Extracting Paddle.tar.gz"
|
|
tar --use-compress-program='pzstd' -xpf Paddle.tar.gz
|
|
rm Paddle.tar.gz
|
|
cd Paddle
|
|
git remote add upstream https://github.com/PaddlePaddle/Paddle.git
|
|
git config pull.rebase false
|
|
git checkout test
|
|
echo "Pull upstream develop"
|
|
source ${{ github.workspace }}/../../../proxy
|
|
bash ci/git_pull.sh $BRANCH
|
|
|
|
- name: Determine the runner
|
|
run: |
|
|
runner_name=`(echo $PWD|awk -F '/' '{print $3}')`
|
|
echo $runner_name
|
|
source ${{ github.workspace }}/Paddle/ci/utils.sh
|
|
determine_npu_runner ${runner_name}
|
|
echo no_proxy="localhost,127.0.0.1,localaddress,.bj.bcebos.com,.localdomain.com,.cdn.bcebos.com,.baidu.com,.bcebos.com" >> ${{ github.env }}
|
|
|
|
- name: Clone PaddleCustomDevice repository
|
|
run: |
|
|
source ~/.bashrc
|
|
git clone --depth=1000 -b develop https://github.com/PaddlePaddle/PaddleCustomDevice.git
|
|
cd PaddleCustomDevice
|
|
cp -r ../Paddle .
|
|
|
|
- name: Check docker image and run container
|
|
env:
|
|
PADDLE_VERSION: 0.0.0
|
|
WITH_COVERAGE: "OFF"
|
|
GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
|
AGILE_PULL_ID: ${{ github.event.pull_request.number }}
|
|
AGILE_REVISION: ${{ github.event.pull_request.head.sha }}
|
|
GIT_PR_ID: ${{ github.event.pull_request.number }}
|
|
PYTHON_VERSION: "3.10"
|
|
no_proxy: bcebos.com
|
|
USE_910B: 1
|
|
FLAGS_set_to_1d: "False"
|
|
NVIDIA_TF32_OVERRIDE: 0
|
|
paddle_submodule: "ON"
|
|
MC2: 1
|
|
HCCL_OP_BASE_FFTS_MODE_ENABLE: "TRUE"
|
|
PADDLE_XCCL_BACKEND: npu
|
|
HCCL_SOCKET_IFNAME: =xgbe0
|
|
FLAGS_eager_communication_connection: 1
|
|
FLAGS_use_stride_kernel: 0
|
|
FLAGS_allocator_strategy: naive_best_fit
|
|
FLAGS_npu_storage_format: 0
|
|
TEST_IMPORTANT: "ON"
|
|
PADDLE_BRANCH: ${{ github.event.pull_request.base.ref }}
|
|
home_dir: ${{ github.workspace }}/../../../..
|
|
run: |
|
|
echo ${ASCEND_RT_VISIBLE_DEVICES}
|
|
container_name=${TASK}-$(date +%Y%m%d-%H%M%S)
|
|
echo "container_name=${container_name}" >> ${{ github.env }}
|
|
docker run --privileged --pids-limit 409600 --shm-size=128G -d -t \
|
|
--cap-add=SYS_PTRACE --security-opt seccomp=unconfined --name ${container_name} \
|
|
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
|
|
-v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
|
|
-v /usr/local/dcmi:/usr/local/dcmi \
|
|
-v /ssd2/workspace/npu-dev/.cache:/root/.cache \
|
|
-v /ssd2/workspace/npu-dev/.ccache:/root/.ccache \
|
|
-v $home_dir/actions-runner:$home_dir/actions-runner \
|
|
-v ${{ github.workspace }}/PaddleCustomDevice:/paddle \
|
|
-e BRANCH \
|
|
-e PR_ID \
|
|
-e COMMIT_ID \
|
|
-e work_dir \
|
|
-e PADDLE_ROOT \
|
|
-e PADDLE_VERSION \
|
|
-e ASCEND_RT_VISIBLE_DEVICES \
|
|
-e WITH_COVERAGE \
|
|
-e GITHUB_API_TOKEN \
|
|
-e AGILE_PULL_ID \
|
|
-e AGILE_REVISION \
|
|
-e GIT_PR_ID \
|
|
-e PYTHON_VERSION \
|
|
-e no_proxy \
|
|
-e USE_910B \
|
|
-e FLAGS_set_to_1d \
|
|
-e NVIDIA_TF32_OVERRIDE \
|
|
-e paddle_submodule \
|
|
-e MC2 \
|
|
-e HCCL_OP_BASE_FFTS_MODE_ENABLE \
|
|
-e PADDLE_XCCL_BACKEND \
|
|
-e HCCL_SOCKET_IFNAME \
|
|
-e FLAGS_eager_communication_connection \
|
|
-e FLAGS_use_stride_kernel \
|
|
-e FLAGS_allocator_strategy \
|
|
-e FLAGS_npu_storage_format \
|
|
-e TEST_IMPORTANT \
|
|
-e PADDLE_BRANCH \
|
|
-e CI_name \
|
|
-w /paddle --network host ${docker_image} /bin/bash
|
|
|
|
- name: Install Paddle-CPU
|
|
run: |
|
|
docker exec -t ${{ env.container_name }} /bin/bash -c '
|
|
set -x
|
|
source ~/.bashrc
|
|
wget -q https://sys-p0.bj.bcebos.com/libstdc%2B%2B6_13.1.0-8ubuntu1_20.04.2_amd64.deb
|
|
dpkg -i *deb
|
|
wget -q --no-proxy https://paddle-github-action.bj.bcebos.com/PR/cpu_whl/${PR_ID}/${COMMIT_ID}/paddlepaddle-0.0.0-cp310-cp310-linux_x86_64.whl --no-check-certificate
|
|
PATH=/usr/local/bin:${PATH}
|
|
echo "export PATH=$PATH" >> ~/.bashrc
|
|
ln -sf $(which python3.10) /usr/local/bin/python
|
|
ln -sf $(which pip3.10) /usr/local/bin/pip
|
|
ln -sf $(which python3.10) /usr/bin/python
|
|
ln -sf $(which pip3.10) /usr/bin/pip
|
|
echo "::group::Install Paddle"
|
|
pip install paddlepaddle-*.whl && rm -rf paddlepaddle-*.whl
|
|
python -c "import paddle; print(paddle.__version__)"
|
|
python -c "import paddle; print(paddle.version.commit)"
|
|
echo "::endgroup::"
|
|
'
|
|
|
|
- name: Build and test
|
|
run: |
|
|
docker exec -t ${{ env.container_name }} /bin/bash -c '
|
|
source ~/.bashrc
|
|
set -x
|
|
echo "::group::Install dependencies"
|
|
python -m pip install PyGithub
|
|
python -m pip install wheel
|
|
echo "::endgroup::"
|
|
source /usr/local/Ascend/ascend-toolkit/set_env.sh
|
|
pip install -U numpy==1.26.4
|
|
git config --global --add safe.directory ${work_dir}
|
|
git config --global --add safe.directory ${work_dir}/Paddle
|
|
cd Paddle
|
|
git submodule foreach "git config --global --add safe.directory \$toplevel/\$sm_path"
|
|
cd ..
|
|
bash backends/npu/tools/pr_ci_npu.sh
|
|
'
|
|
|
|
- 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}
|