name: CI-H on: pull_request: types: [opened, synchronize] branches: [develop, release/**] permissions: read-all concurrency: group: ${{ github.event.pull_request.number }}-${{ github.workflow }} cancel-in-progress: true env: PR_ID: ${{ github.event.pull_request.number }} COMMIT_ID: ${{ github.event.pull_request.head.sha }} TASK: paddle-CI-${{ github.event.pull_request.number }}-coverage ci_scripts: /paddle/ci BRANCH: ${{ github.base_ref }} work_dir: /paddle PADDLE_ROOT: /paddle GIT_PR_ID: ${{ github.event.pull_request.number }} CI_name: h-coverage CFS_DIR: /home/data/cfs no_proxy: "bcebos.com,apiin.im.baidu.com,gitee.com,aliyun.com,.baidu.com,.tuna.tsinghua.edu.cn" fleet_branch: "release/0.2" defaults: run: shell: bash jobs: check-skill: name: Check file paths if: ${{ github.repository_owner == 'PaddlePaddle' }} runs-on: group: APPROVAL continue-on-error: true permissions: contents: read actions: read outputs: can-skip: ${{ steps.skip-skill.outputs.should-skip }} steps: - name: Cleanup run: | rm -rf * .[^.]* - name: Checkout uses: actions/checkout@v6 - name: Run skip check id: skip-skill uses: ./.github/actions/skip-skill with: ignore-paths: '.agents/skills/**' - name: Print outputs if: always() run: | echo "=== check-skill outputs ===" echo "steps.skip-skill.outputs.should-skip: '${{ steps.skip-skill.outputs.should-skip }}'" echo "============================" clone: name: Coverage clone needs: [check-skill] uses: ./.github/workflows/_Clone-linux.yml with: workflow-name: "coverage" clone_dir: h-ci skill-can-skip: ${{ needs.check-skill.outputs.can-skip }} check-bypass: name: Check bypass needs: [clone] if: ${{ needs.clone.outputs.can-skip != 'true' }} uses: ./.github/workflows/check-bypass.yml with: workflow-name: h-ci secrets: github-token: ${{ secrets.GITHUB_TOKEN }} build: name: Coverage build needs: [clone, check-bypass] if: ${{ needs.check-bypass.outputs.can-skip != 'true' && needs.clone.outputs.can-skip != 'true' }} runs-on: group: GZ_BD-CPU timeout-minutes: 120 outputs: can-skip: ${{ needs.check-bypass.outputs.can-skip == 'true' || needs.clone.outputs.can-skip == 'true' || 'false' }} steps: - name: Check docker image and run container env: CACHE_DIR: "/root/.cache/coverage" 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 FLAGS_fraction_of_gpu_memory_to_use: 0.15 CTEST_PARALLEL_LEVEL: 2 WITH_GPU: "ON" CUDA_ARCH_NAME: Hopper WITH_AVX: "ON" PADDLE_VERSION: 0.0.0 CUDA_VISIBLE_DEVICES: 0,1 WITH_DISTRIBUTE: "ON" LITE_GIT_TAG: develop WITH_UNITY_BUILD: "ON" WITH_FA_BUILD_WITH_CACHE: "ON" PY_VERSION: "3.12" INFERENCE_DEMO_INSTALL_DIR: /root/.cache/coverage CCACHE_MAXSIZE: 50G CCACHE_LIMIT_MULTIPLE: 0.8 GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }} run: | container_name=${TASK}-build-$(date +%Y%m%d-%H%M%S) echo "container_name=${container_name}" >> ${{ github.env }} docker_image=ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:ubuntu24-cuda129-py312-dev docker run -d -t --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 ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \ -v ${{ github.workspace }}:/paddle \ -e CI_name \ -e BRANCH \ -e PR_ID \ -e COMMIT_ID \ -e work_dir \ -e PADDLE_ROOT \ -e GIT_PR_ID \ -e CACHE_DIR \ -e CCACHE_DIR \ -e CCACHE_SECONDARY_STORAGE \ -e ci_scripts \ -e FLAGS_fraction_of_gpu_memory_to_use \ -e CTEST_PARALLEL_LEVEL \ -e WITH_GPU \ -e CUDA_ARCH_NAME \ -e WITH_AVX \ -e PADDLE_VERSION \ -e WITH_DISTRIBUTE \ -e LITE_GIT_TAG \ -e WITH_UNITY_BUILD \ -e WITH_FA_BUILD_WITH_CACHE \ -e PY_VERSION \ -e INFERENCE_DEMO_INSTALL_DIR \ -e CCACHE_MAXSIZE \ -e CCACHE_LIMIT_MULTIPLE \ -e GITHUB_TOKEN \ -e GITHUB_API_TOKEN \ -e CFS_DIR \ -e no_proxy \ -w /paddle --network host ${docker_image} - name: Download paddle.tar.gz and update test branch timeout-minutes: 30 run: | docker exec -t ${{ env.container_name }} /bin/bash -c ' rm -rf * .[^.]* set -e echo "Downloading Paddle.tar.gz" wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PR/h-ci/${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 --add safe.directory "*" git remote -v set +e git remote add upstream https://github.com/PaddlePaddle/Paddle.git set -e git config pull.rebase false git checkout test echo "Pull upstream $BRANCH" source ${{ github.workspace }}/../../../proxy bash ci/git_pull.sh $BRANCH git submodule update ' - name: Build run: | docker exec -t ${{ env.container_name }} /bin/bash -c ' flashattn_version=$(git submodule status|grep flashattn|awk "{print \$1}"|sed "s#-##g") echo flashattn_version:$flashattn_version url="https://xly-devops.bj.bcebos.com/gpups/flash-attention/cu90/flashattn_libs_${flashattn_version}.tar" echo url:$url url_return=`curl -s -o /dev/null -w "%{http_code}" $url` if [ "$url_return" != "200" ];then echo "flashattn cache not found, please contact umiswing" exit 7 fi mkdir -p ${CFS_DIR}/.cache/coverage mkdir -p ${CFS_DIR}/.ccache/coverage export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-12.9/compat source ${{ github.workspace }}/../../../proxy bash ${ci_scripts}/cmake-predownload.sh bash ${ci_scripts}/cmake_preload_data_file.sh pip install -r python/requirements.txt mkdir -p build && cd build ccache -z cmake .. -DPY_VERSION=3.12 -DWITH_GPU=ON -DWITH_DISTRIBUTE=ON -DWITH_TESTING=ON -DCUDA_ARCH_NAME=Manual -DCUDA_ARCH_BIN="90" -DFA_JOB_POOLS_COMPILE=1 -DWITH_CUDNN_FRONTEND=ON -DON_INFER=OFF -DWITH_NVSHMEM=ON make -j20 EXIT_CODE=$? ccache -s exit $EXIT_CODE ' - name: Clean up env run: | docker exec -t ${{ env.container_name }} /bin/bash -c ' source ~/.bashrc source ${ci_scripts}/utils.sh; clean_build_files : "${PY_VERSION:?PY_VERSION is required}" rm -rf $(find . -name "*.a") rm -rf $(find . -name "*.o") rm -rf lib.linux-x86_64-${PY_VERSION} find ./ -name "eager_generator" -or -name "kernel_signature_generator" -or -name "eager_legacy_op_function_generator" | xargs rm -rf rm -rf ./python/build/lib.linux-x86_64-${PY_VERSION}/ cd "${work_dir}/build/third_party" && find $(ls | grep -v "dlpack" | grep -v "install" | grep -v "eigen3" | grep -v "gflags") -type f ! -name "*.so" -a ! -name "libdnnl.so*" -delete cd / tar --use-compress-program="pzstd -1" -cf Paddle.tar.gz paddle ' - name: Upload coverage product env: home_path: ${{ github.workspace }}/.. bos_file: ${{ github.workspace }}/../bos_retry/BosClient.py paddle_whl: paddlepaddle_gpu-0.0.0-cp312-cp312-linux_x86_64.whl run: | docker exec -t ${{ env.container_name }} /bin/bash -c ' echo "::group::Install bce-python-sdk" python -m pip install bce-python-sdk==0.8.74 echo "::endgroup::" export AK=paddle export SK=paddle if [ ! -f "${{ env.bos_file }}" ]; then wget -q --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 cd /paddle mv /Paddle.tar.gz . cp ./build/python/dist/paddlepaddle_gpu-0.0.0-cp312-cp312-linux_x86_64.whl . echo "Uploading Paddle.tar.gz" python ${{ env.bos_file }} Paddle.tar.gz paddle-github-action/PR/h-coverage/${{ env.PR_ID }}/${{ env.COMMIT_ID }} echo "Uploading coverage wheel" python ${{ env.bos_file }} ${{ env.paddle_whl }} paddle-github-action/PR/h-coverage/${{ env.PR_ID }}/${{ env.COMMIT_ID }} echo "End Upload" ' - name: Terminate and delete the container if: always() run: | set +e docker exec -t ${{ env.container_name }} /bin/bash -c 'rm -rf * .[^.]*' docker stop ${{ env.container_name }} docker rm ${{ env.container_name }} test: name: Coverage test needs: [build, check-bypass, check-skill] if: ${{ needs.build.outputs.can-skip != 'true' && needs.check-bypass.outputs.can-skip != 'true' && needs.check-skill.outputs.can-skip != 'true' }} runs-on: group: H-Coverage timeout-minutes: 120 steps: - name: Determine the runner run: | runner_name=`(echo $PWD|awk -F '/' '{print $3}')` echo $runner_name wget -q https://xly-devops.bj.bcebos.com/utils.sh source utils.sh determine_gpu_runner ${runner_name} - name: Check docker image and run container env: CACHE_DIR: "/root/.cache/coverage" CCACHE_DIR: "/root/.ccache/coverage" FLAGS_fraction_of_gpu_memory_to_use: 0.15 CTEST_PARALLEL_LEVEL: 2 WITH_GPU: "ON" CUDA_ARCH_NAME: Hopper WITH_AVX: "ON" COVERALLS_UPLOAD: "ON" PADDLE_VERSION: 0.0.0 WITH_DISTRIBUTE: "ON" WITH_UNITY_BUILD: "ON" PY_VERSION: "3.12" WITH_SHARED_PHI: "ON" GPU_DEVICES: ${{ env.GPU_DEVICES }} WITH_CINN: "ON" INFERENCE_DEMO_INSTALL_DIR: /root/.cache/coverage CCACHE_MAXSIZE: 200G CCACHE_LIMIT_MULTIPLE: 0.8 FLAGS_PIR_OPTEST: "TRUE" ON_INFER: "ON" COVERAGE_FILE: ${{ github.workspace }}/build/python-coverage.data GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }} run: | container_name=${TASK}-$(date +%Y%m%d-%H%M%S) echo "container_name=${container_name}" >> ${{ github.env }} docker_image=ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:ubuntu24-cuda129-py312-dev docker run -d -t --gpus "\"device=${GPU_DEVICES}\"" --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 "/dev/shm:/dev/shm" \ -v ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \ -v ${{ github.workspace }}:/paddle \ -e CI_name \ -e BRANCH \ -e PR_ID \ -e COMMIT_ID \ -e work_dir \ -e PADDLE_ROOT \ -e GIT_PR_ID \ -e CACHE_DIR \ -e CCACHE_DIR \ -e ci_scripts \ -e FLAGS_fraction_of_gpu_memory_to_use \ -e CTEST_PARALLEL_LEVEL \ -e WITH_GPU \ -e CUDA_ARCH_NAME \ -e WITH_AVX \ -e WITH_COVERAGE \ -e COVERALLS_UPLOAD \ -e PADDLE_VERSION \ -e WITH_DISTRIBUTE \ -e WITH_UNITY_BUILD \ -e PY_VERSION \ -e WITH_SHARED_PHI \ -e WITH_CINN \ -e INFERENCE_DEMO_INSTALL_DIR \ -e CCACHE_MAXSIZE \ -e CCACHE_LIMIT_MULTIPLE \ -e FLAGS_PIR_OPTEST \ -e ON_INFER \ -e COVERAGE_FILE \ -e GITHUB_TOKEN \ -e GITHUB_API_TOKEN \ -e CFS_DIR \ -e no_proxy \ -w /paddle --network host ${docker_image} - name: Download paddle.tar.gz and update test branch timeout-minutes: 30 run: | docker exec -t ${{ env.container_name }} /bin/bash -c ' rm -rf * .[^.]* set -e echo "Downloading Paddle.tar.gz from cfs" wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PR/h-coverage/${PR_ID}/${COMMIT_ID}/Paddle.tar.gz --no-check-certificate echo "Extracting Paddle.tar.gz" tar --use-compress-program="pzstd -1" -xf Paddle.tar.gz --strip-components=1 rm Paddle.tar.gz ' - name: Test id: unit_test run: | docker exec -t ${{ env.container_name }} /bin/bash -c ' source ${{ github.workspace }}/../../../proxy pip install build//python/dist/*.whl --no-deps pip install -r python/unittest_py/requirements.txt bash $ci_scripts/h-test.sh ' - name: FA Test if: (success() || failure()) && steps.unit_test.conclusion != 'skipped' run: | docker exec -t ${{ env.container_name }} /bin/bash -c ' source ${{ github.workspace }}/../../../proxy python -m pip install pytest einops cd test/test_flashmask_ci bash run.sh ' - name: Terminate and delete the container if: always() run: | set +e rm Paddle.tar.gz docker exec -t ${{ env.container_name }} /bin/bash -c 'rm -rf * .[^.]*' docker stop ${{ env.container_name }} docker rm ${{ env.container_name }} fleet-build: name: Build Fleet whl needs: [build, check-bypass, check-skill] if: ${{ needs.build.outputs.can-skip != 'true' && needs.check-bypass.outputs.can-skip != 'true' && needs.check-skill.outputs.can-skip != 'true' }} runs-on: group: GZ_BD-CPU timeout-minutes: 60 env: TASK: fleet-ci-paddle-build-whl-${{ github.event.pull_request.number }} docker_image: "ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:ubuntu24-cuda129-py312-dev" PY_VERSION: "3.12" steps: - name: Check docker image and run container env: 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 CCACHE_STATSLOG: /paddle/build/.stats.log CCACHE_SLOPPINESS: clang_index_store,time_macros,include_file_mtime 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 --name ${container_name} \ -v "/dev/shm:/dev/shm" \ -v "/home/data/cfs:/home/data/cfs" \ -v "/home/data/shared:/home/data/shared" \ -v ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \ -v ${{ github.workspace }}:/paddle \ -e BRANCH \ -e PR_ID \ -e COMMIT_ID \ -e PADDLE_ROOT \ -e ci_scripts \ -e no_proxy \ -e fleet_branch \ -e CI_name \ -e CCACHE_DIR \ -e CCACHE_SECONDARY_STORAGE \ -e CCACHE_MAXSIZE \ -e CCACHE_LIMIT_MULTIPLE \ -e CCACHE_STATSLOG \ -e CCACHE_SLOPPINESS \ -e PY_VERSION \ -w /paddle --network host ${docker_image} - name: uv build whl run: | docker exec -t ${{ env.container_name }} /bin/bash -ce ' rm -rf * .[^.]* set -x source ${{ github.workspace }}/../../../proxy apt-get update apt-get install -y libnvidia-compute-535 python -m pip install --upgrade pip wheel setuptools git clone https://github.com/PaddlePaddle/PaddleFleet.git . git config --global --add safe.directory /paddle git config user.name "PaddleCI" git config user.email "paddle_ci@example.com" git config pull.rebase false if [ "${BRANCH}" != "develop" ]; then git checkout $fleet_branch echo "Checked out fleet branch: $fleet_branch" else git checkout develop echo "Checked out fleet branch: develop" fi mkdir -p /home/.cache/pip pip cache dir echo "Install uv" export UV_HTTP_TIMEOUT=300 pip install uv coverage==7.6.1 bce-python-sdk==0.8.74 wrapt echo "uv build" git submodule update --init --recursive export LD_LIBRARY_PATH=/usr/local/cuda-12.9/compat:$LD_LIBRARY_PATH export IS_NVIDIA=True echo "Downloading Paddle.tar.gz from cfs" wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PR/h-coverage/${PR_ID}/${COMMIT_ID}/paddlepaddle_gpu-0.0.0-cp312-cp312-linux_x86_64.whl --no-check-certificate pip uninstall paddlepaddle-gpu -y pip install paddlepaddle_gpu-0.0.0-cp312-cp312-linux_x86_64.whl --force-reinstall --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/ pip install paddle-nvidia-nvshmem-cu12 --extra-index-url https://www.paddlepaddle.org.cn/packages/nightly/cu129/ uv build --wheel --no-build-isolation -v --python $(which python) ' - name: download ops whl from latest run: | docker exec -t ${{ env.container_name }} /bin/bash -ce ' set -x OPS_DST=paddlefleet_ops-0.0.0-cp312-cp312-linux_x86_64.whl wget -q --no-proxy "https://paddle-whl.bj.bcebos.com/nightly/cu129/paddlefleet-ops/${OPS_DST}" -O dist/${OPS_DST} ' - name: upload whl to BOS run: | docker exec -t ${{ env.container_name }} /bin/bash -ce ' export AK=paddle export SK=paddle cd / if [ ! -f "bos/BosClient.py" ]; then wget -q --no-proxy -O bos_new.tar.gz https://xly-devops.bj.bcebos.com/home/bos_new.tar.gz --no-check-certificate mkdir bos tar xf bos_new.tar.gz -C bos fi target_path="paddle-github-action/PR/paddlefleet/${{ env.PR_ID }}/${{ env.COMMIT_ID }}" tar zcf paddldfleet.tar.gz paddle python /bos/BosClient.py paddldfleet.tar.gz ${target_path} ' - name: Terminate and delete the container if: always() run: | set +e docker exec -t ${{ env.container_name }} /bin/bash -c 'bash ci/clean_uv_cache.sh; rm -rf * .[^.]*' docker rm -f ${{ env.container_name }} fleet_single_card_test: name: Fleet Unit test (single card) needs: [build, fleet-build] if: needs.build.outputs.can-skip != 'true' runs-on: group: Fleet-H-single-card timeout-minutes: 90 env: PIP_CACHE_DIR: /root/.cache/pip CACHE_DIR: /root/.cache TASK: paddle-fleet-CI-${{ github.event.pull_request.number }}-single-card-test PY_VERSION: "3.12" steps: - name: Determine the runner run: | gpu_id=$(( $(echo $PWD | awk -F'/' '{print $3}' | awk -F'-' '{print $2}') - 1 )) echo GPU_DEVICES="$gpu_id" >> $GITHUB_ENV - name: Check docker image and run container env: GPU_DEVICES: ${{ env.GPU_DEVICES }} docker_image: "ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:ubuntu24-cuda129-py312-dev" run: | container_name=${TASK}-$(date +%Y%m%d-%H%M%S) echo "container_name=${container_name}" >> ${{ github.env }} docker pull $docker_image docker run -d -t --name ${container_name} --gpus "\"device=${GPU_DEVICES}\"" --shm-size=32G \ -v "/dev/shm:/dev/shm" \ -v ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \ -v ${{ github.workspace }}/../../../proxy:/root/proxy \ -v /ssd1/paddle-1/action_cache:/root/.cache \ -v ${{ github.workspace }}:/paddle \ -e BRANCH \ -e PR_ID \ -e COMMIT_ID \ -e PADDLE_ROOT \ -e ci_scripts \ -e CACHE_DIR \ -e no_proxy \ -e CI_name \ -e PIP_CACHE_DIR \ -e PY_VERSION \ -e work_dir \ -e GITHUB_SHA="${{ github.event.pull_request.head.sha }}" \ -e GITHUB_HEAD_REF="${{ github.head_ref }}" \ -e GITHUB_BASE_SHA="${{ github.event.pull_request.base.sha }}" \ -e GITHUB_REPO_NAME="${{ github.repository }}" \ -e GITHUB_EVENT_PULL_REQUEST_NUMBER="${{ github.event.pull_request.number }}" \ -e GITHUB_TOKEN="${{ secrets.GITHUB_TOKEN }}" \ -e GITHUB_RUN_ID="${{ github.run_id }}" \ -w /paddle --network host ${docker_image} - name: Install PaddleFleet run: | docker exec -t ${{ env.container_name }} /bin/bash -ce ' rm -rf * .[^.]* source /root/proxy export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64 pip install uv coverage==7.6.1 bce-python-sdk==0.8.74 wrapt pytest matplotlib parameterized wget -q --tries=5 --no-proxy --no-check-certificate https://paddle-github-action.cdn.bcebos.com/PR/paddlefleet/${PR_ID}/${COMMIT_ID}/paddldfleet.tar.gz tar -xf paddldfleet.tar.gz --strip-components=1 git config --global --add safe.directory /paddle pip install dist/paddlefleet-*.whl --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/ --extra-index-url=https://www.paddlepaddle.org.cn/packages/stable/cu129/ pip install dist/paddlefleet_ops-*.whl --extra-index-url=https://www.paddlepaddle.org.cn/packages/stable/cu129/ --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/ wget https://github.com/mikefarah/yq/releases/latest/download/yq_linux_amd64 -O /usr/local/bin/yq chmod +x /usr/local/bin/yq ' - name: Download paddle.tar.gz and install paddle whl timeout-minutes: 30 run: | docker exec -t ${{ env.container_name }} /bin/bash -c ' set -e export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64 mkdir -p /PaddlePaddle cd /PaddlePaddle echo "Downloading Paddle.tar.gz from cfs" wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PR/h-coverage/${PR_ID}/${COMMIT_ID}/paddlepaddle_gpu-0.0.0-cp312-cp312-linux_x86_64.whl --no-check-certificate export UV_SKIP_WHEEL_FILENAME_CHECK=1 #This environment variable allows installing the latest commit-level whl package of Paddle. export UV_NO_SYNC=1 # This environment variable prevents uv sync from being executed when running un run. export UV_HTTP_TIMEOUT=300 pip uninstall paddlepaddle-gpu -y pip install paddlepaddle_gpu-0.0.0-cp312-cp312-linux_x86_64.whl --force-reinstall --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/ echo "paddlefleet commit:" python -c "import paddlefleet; print(paddlefleet.version.commit)" ' - name: Single card test timeout-minutes: 60 run: | docker exec -t ${{ env.container_name }} /bin/bash -xce ' pwd export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64 if [ "${BRANCH}" != "develop" ]; then git checkout $fleet_branch echo "Checked out fleet branch: $fleet_branch" else git checkout develop echo "Checked out fleet branch: develop" fi export UV_SKIP_WHEEL_FILENAME_CHECK=1 #This environment variable allows installing the latest commit-level whl package of Paddle. export UV_NO_SYNC=1 # This environment variable prevents uv sync from being executed when running un run. export UV_HTTP_TIMEOUT=300 python -c "import paddle; print(paddle.version.commit)" bash ci/single_card_test.sh single_card_exit_code=$? if [[ "$single_card_exit_code" != "0" ]]; then echo -e "::error:: \033[31mSingle card test failed.\033[0m" exit 1 else echo -e "\033[32mSingle card test succeeded.\033[0m" fi ' - name: Terminate and delete the container if: always() run: | set +e docker exec -t ${{ env.container_name }} /bin/bash -c 'rm -rf * .[^.]*' docker rm -f ${{ env.container_name }} fleet-multi-card_test: name: Fleet Unit test (multi-card) needs: [build, fleet-build, check-bypass, check-skill] if: ${{ needs.build.outputs.can-skip != 'true' && needs.check-bypass.outputs.can-skip != 'true' && needs.check-skill.outputs.can-skip != 'true' }} runs-on: group: Fleet-H-multi-card timeout-minutes: 60 env: PIP_CACHE_DIR: /root/.cache/pip TASK: paddle-fleet-CI-${{ github.event.pull_request.number }}-multi-card_test docker_image: "ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:ubuntu24-cuda129-py312-dev" PY_VERSION: "3.12" steps: - name: Check docker image and run container 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 "/dev/shm:/dev/shm" \ -v ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \ -v ${{ github.workspace }}/../../../proxy:/root/proxy \ -v ${{ github.workspace }}/../../../.cache:/root/.cache \ -v ${{ github.workspace }}:/paddle \ -e BRANCH \ -e PR_ID \ -e COMMIT_ID \ -e PADDLE_ROOT \ -e ci_scripts \ -e CACHE_DIR \ -e fleet_branch \ -e no_proxy \ -e CI_name \ -e PIP_CACHE_DIR \ -e PY_VERSION \ -e work_dir \ -e GITHUB_SHA="${{ github.event.pull_request.head.sha }}" \ -e GITHUB_HEAD_REF="${{ github.head_ref }}" \ -e GITHUB_BASE_SHA="${{ github.event.pull_request.base.sha }}" \ -e GITHUB_REPO_NAME="${{ github.repository }}" \ -e GITHUB_EVENT_NAME="${{ github.event_name }}" \ -e GITHUB_EVENT_PULL_REQUEST_NUMBER="${{ github.event.pull_request.number }}" \ -e GITHUB_TOKEN="${{ secrets.GITHUB_TOKEN }}" \ -e GITHUB_RUN_ID="${{ github.run_id }}" \ -w /paddle --network host ${docker_image} - name: Install PaddleFleet run: | docker exec -t ${{ env.container_name }} /bin/bash -ce ' rm -rf * .[^.]* source /root/proxy export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64 pip install uv coverage==7.6.1 bce-python-sdk==0.8.74 wrapt pytest matplotlib parameterized wget -q --tries=5 --no-proxy --no-check-certificate https://paddle-github-action.cdn.bcebos.com/PR/paddlefleet/${PR_ID}/${COMMIT_ID}/paddldfleet.tar.gz tar -xf paddldfleet.tar.gz --strip-components=1 git config --global --add safe.directory /paddle pip install dist/paddlefleet-*.whl --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/ --extra-index-url=https://www.paddlepaddle.org.cn/packages/stable/cu129/ pip install dist/paddlefleet_ops-*.whl --extra-index-url=https://www.paddlepaddle.org.cn/packages/stable/cu129/ --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/ wget https://github.com/mikefarah/yq/releases/latest/download/yq_linux_amd64 -O /usr/local/bin/yq chmod +x /usr/local/bin/yq ' - name: Download paddle.tar.gz and install paddle whl timeout-minutes: 30 run: | docker exec -t ${{ env.container_name }} /bin/bash -c ' set -e export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64 mkdir -p /PaddlePaddle cd /PaddlePaddle echo "Downloading Paddle.tar.gz from cfs" wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PR/h-coverage/${PR_ID}/${COMMIT_ID}/paddlepaddle_gpu-0.0.0-cp312-cp312-linux_x86_64.whl --no-check-certificate export UV_SKIP_WHEEL_FILENAME_CHECK=1 #This environment variable allows installing the latest commit-level whl package of Paddle. export UV_NO_SYNC=1 # This environment variable prevents uv sync from being executed when running un run. export UV_HTTP_TIMEOUT=300 pip uninstall paddlepaddle-gpu -y pip install paddlepaddle_gpu-0.0.0-cp312-cp312-linux_x86_64.whl --force-reinstall --extra-index-url=https://www.paddlepaddle.org.cn/packages/nightly/cu129/ echo "paddlefleet commit:" python -c "import paddlefleet; print(paddlefleet.version.commit)" ' - name: Multi-card test timeout-minutes: 60 run: | docker exec -t ${{ env.container_name }} /bin/bash -ce ' export PYTHONPATH=$(pwd) export LD_LIBRARY_PATH=/usr/local/cuda-12.9/targets/x86_64-linux/lib:/usr/local/cuda/lib64 if [ "${BRANCH}" != "develop" ]; then git checkout $fleet_branch echo "Checked out fleet branch: $fleet_branch" else git checkout develop echo "Checked out fleet branch: develop" fi python -c "import paddle; print(paddle.version.commit)" export UV_SKIP_WHEEL_FILENAME_CHECK=1 #This environment variable allows installing the latest commit-level whl package of Paddle. export UV_NO_SYNC=1 # This environment variable prevents uv sync from being executed when running un run. export UV_HTTP_TIMEOUT=300 bash ci/multi-card_test.sh multi_card_exit_code=$? if [[ "$multi_card_exit_code" != "0" ]]; then echo -e "::error:: \033[31mMulti card test failed.\033[0m" exit 1 else echo -e "\033[32mMulti card test succeeded.\033[0m" fi ' - name: Terminate and delete the container if: always() run: | set +e docker exec -t ${{ env.container_name }} /bin/bash -c 'rm -rf * .[^.]*' docker rm -f ${{ env.container_name }}