|
| 1 | +name: CI-H-Coverage |
| 2 | + |
| 3 | +on: |
| 4 | + pull_request: |
| 5 | + types: [opened, synchronize] |
| 6 | + branches: [develop, release/**] |
| 7 | + |
| 8 | +permissions: read-all |
| 9 | + |
| 10 | +concurrency: |
| 11 | + group: ${{ github.event.pull_request.number }}-${{ github.workflow }} |
| 12 | + cancel-in-progress: true |
| 13 | + |
| 14 | +env: |
| 15 | + PR_ID: ${{ github.event.pull_request.number }} |
| 16 | + COMMIT_ID: ${{ github.event.pull_request.head.sha }} |
| 17 | + TASK: paddle-CI-${{ github.event.pull_request.number }}-coverage |
| 18 | + ci_scripts: /paddle/ci |
| 19 | + BRANCH: ${{ github.base_ref }} |
| 20 | + work_dir: /paddle |
| 21 | + PADDLE_ROOT: /paddle |
| 22 | + GIT_PR_ID: ${{ github.event.pull_request.number }} |
| 23 | + CI_name: h-coverage |
| 24 | + CFS_DIR: /home/data/cfs |
| 25 | + no_proxy: "bcebos.com,apiin.im.baidu.com,gitee.com,aliyun.com,.baidu.com,.tuna.tsinghua.edu.cn" |
| 26 | + |
| 27 | +defaults: |
| 28 | + run: |
| 29 | + shell: bash |
| 30 | + |
| 31 | +jobs: |
| 32 | + clone: |
| 33 | + name: Coverage clone |
| 34 | + uses: ./.github/workflows/_Clone-linux.yml |
| 35 | + with: |
| 36 | + workflow-name: 'coverage' |
| 37 | + clone_dir: Paddle-coverage |
| 38 | + |
| 39 | + build: |
| 40 | + name: Coverage build |
| 41 | + needs: [clone] |
| 42 | + if: needs.clone.outputs.can-skip != 'true' |
| 43 | + runs-on: |
| 44 | + group: GZ_BD-CPU |
| 45 | + outputs: |
| 46 | + can-skip: ${{ steps.check-bypass.outputs.can-skip }} |
| 47 | + |
| 48 | + steps: |
| 49 | + - name: Check docker image and run container |
| 50 | + env: |
| 51 | + CACHE_DIR: "/root/.cache/coverage" |
| 52 | + CCACHE_DIR: "/root/.ccache/h-coverage" |
| 53 | + FLAGS_fraction_of_gpu_memory_to_use: 0.15 |
| 54 | + CTEST_PARALLEL_LEVEL: 2 |
| 55 | + WITH_GPU: "ON" |
| 56 | + CUDA_ARCH_NAME: Hopper |
| 57 | + WITH_AVX: "ON" |
| 58 | + PADDLE_VERSION: 0.0.0 |
| 59 | + CUDA_VISIBLE_DEVICES: 0,1 |
| 60 | + WITH_DISTRIBUTE: "ON" |
| 61 | + LITE_GIT_TAG: develop |
| 62 | + WITH_UNITY_BUILD: "ON" |
| 63 | + WITH_FA_BUILD_WITH_CACHE: "ON" |
| 64 | + PY_VERSION: "3.10" |
| 65 | + INFERENCE_DEMO_INSTALL_DIR: /root/.cache/coverage |
| 66 | + CCACHE_MAXSIZE: 200G |
| 67 | + CCACHE_LIMIT_MULTIPLE: 0.8 |
| 68 | + GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} |
| 69 | + GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }} |
| 70 | + run: | |
| 71 | + container_name=${TASK}-build-$(date +%Y%m%d-%H%M%S) |
| 72 | + echo "container_name=${container_name}" >> ${{ github.env }} |
| 73 | + docker_image=ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:cuda129-coverage-test |
| 74 | + docker run -d -t --name ${container_name} \ |
| 75 | + -v "/home/data/cfs:/home/data/cfs" \ |
| 76 | + -v "/home/data/cfs/.cache:/root/.cache" \ |
| 77 | + -v "/home/data/cfs/.ccache:/root/.ccache" \ |
| 78 | + -v "/dev/shm:/dev/shm" \ |
| 79 | + -v ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \ |
| 80 | + -v ${{ github.workspace }}:/paddle \ |
| 81 | + -e CI_name \ |
| 82 | + -e BRANCH \ |
| 83 | + -e PR_ID \ |
| 84 | + -e COMMIT_ID \ |
| 85 | + -e work_dir \ |
| 86 | + -e PADDLE_ROOT \ |
| 87 | + -e GIT_PR_ID \ |
| 88 | + -e CACHE_DIR \ |
| 89 | + -e CCACHE_DIR \ |
| 90 | + -e ci_scripts \ |
| 91 | + -e FLAGS_fraction_of_gpu_memory_to_use \ |
| 92 | + -e CTEST_PARALLEL_LEVEL \ |
| 93 | + -e WITH_GPU \ |
| 94 | + -e CUDA_ARCH_NAME \ |
| 95 | + -e WITH_AVX \ |
| 96 | + -e PADDLE_VERSION \ |
| 97 | + -e WITH_DISTRIBUTE \ |
| 98 | + -e LITE_GIT_TAG \ |
| 99 | + -e WITH_UNITY_BUILD \ |
| 100 | + -e WITH_FA_BUILD_WITH_CACHE \ |
| 101 | + -e PY_VERSION \ |
| 102 | + -e INFERENCE_DEMO_INSTALL_DIR \ |
| 103 | + -e CCACHE_MAXSIZE \ |
| 104 | + -e CCACHE_LIMIT_MULTIPLE \ |
| 105 | + -e GITHUB_TOKEN \ |
| 106 | + -e GITHUB_API_TOKEN \ |
| 107 | + -e CFS_DIR \ |
| 108 | + -e no_proxy \ |
| 109 | + -w /paddle --network host ${docker_image} |
| 110 | +
|
| 111 | + - name: Download paddle.tar.gz and update test branch |
| 112 | + run: | |
| 113 | + docker exec -t ${{ env.container_name }} /bin/bash -c ' |
| 114 | + rm -rf * .[^.]* |
| 115 | + set -e |
| 116 | + echo "Downloading Paddle.tar.gz" |
| 117 | + wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PR/Paddle-coverage/${PR_ID}/${COMMIT_ID}/Paddle.tar.gz --no-check-certificate |
| 118 | + echo "Extracting Paddle.tar.gz" |
| 119 | + tar -xf Paddle.tar.gz --strip-components=1 |
| 120 | + rm Paddle.tar.gz |
| 121 | + git config --global --add safe.directory "*" |
| 122 | + git remote -v |
| 123 | + set +e |
| 124 | + git remote add upstream https://github.com/PaddlePaddle/Paddle.git |
| 125 | + set -e |
| 126 | + git config pull.rebase false |
| 127 | + git checkout test |
| 128 | + echo "Pull upstream $BRANCH" |
| 129 | + source ${{ github.workspace }}/../../../proxy |
| 130 | + bash ci/git_pull.sh $BRANCH |
| 131 | + ' |
| 132 | +
|
| 133 | + - name: Check bypass |
| 134 | + id: check-bypass |
| 135 | + uses: ./.github/actions/check-bypass |
| 136 | + with: |
| 137 | + github-token: ${{ secrets.GITHUB_TOKEN }} |
| 138 | + workflow-name: h-ci |
| 139 | + |
| 140 | + - name: Build |
| 141 | + if: steps.check-bypass.outputs.can-skip != 'true' |
| 142 | + run: | |
| 143 | + docker exec -t ${{ env.container_name }} /bin/bash -c ' |
| 144 | + flashattn_version=$(git submodule status|grep flashattn|awk "{print \$1}"|sed "s#-##g") |
| 145 | + url="https://xly-devops.bj.bcebos.com/gpups/flash-attention/cu90/flashattn_libs_${flashattn_version}.tar" |
| 146 | + url_return=`curl -s -o /dev/null -w "%{http_code}" $url` |
| 147 | + if [ "$url_return" != "200" ];then |
| 148 | + echo "flashattn cache not found, please contact umiswing" |
| 149 | + exit 7 |
| 150 | + fi |
| 151 | + mkdir -p ${CFS_DIR}/.cache/coverage |
| 152 | + mkdir -p ${CFS_DIR}/.ccache/coverage |
| 153 | + export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-12.9/compat |
| 154 | + source ${{ github.workspace }}/../../../proxy |
| 155 | + pip install -r python/requirements.txt |
| 156 | + mkdir build && cd build |
| 157 | + cmake .. -DPY_VERSION=3.10 -DWITH_GPU=ON -DWITH_DISTRIBUTE=ON -DWITH_TESTING=ON -DCUDA_ARCH_NAME=Hopper -DFA_JOB_POOLS_COMPILE=1 -DWITH_CUDNN_FRONTEND=ON -DON_INFER=OFF |
| 158 | + make -j20 |
| 159 | + ' |
| 160 | +
|
| 161 | + - name: Clean up env |
| 162 | + if: steps.check-bypass.outputs.can-skip != 'true' |
| 163 | + run: | |
| 164 | + docker exec -t ${{ env.container_name }} /bin/bash -c ' |
| 165 | + source ~/.bashrc |
| 166 | + source ${ci_scripts}/utils.sh; clean_build_files |
| 167 | + rm -rf $(find . -name "*.a") |
| 168 | + rm -rf $(find . -name "*.o") |
| 169 | + rm -rf lib.linux-x86_64-3.9 |
| 170 | + find ./ -name "eager_generator" -or -name "kernel_signature_generator" -or -name "eager_legacy_op_function_generator" | xargs rm -rf |
| 171 | + rm -rf ./python/build/lib.linux-x86_64-3.9/ |
| 172 | + cd "${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 |
| 173 | + cd / |
| 174 | + tar --use-compress-program="pzstd -1" -cf Paddle.tar.gz paddle |
| 175 | + ' |
| 176 | +
|
| 177 | + - name: Upload coverage product |
| 178 | + if: steps.check-bypass.outputs.can-skip != 'true' |
| 179 | + env: |
| 180 | + home_path: ${{ github.workspace }}/.. |
| 181 | + bos_file: ${{ github.workspace }}/../bos_retry/BosClient.py |
| 182 | + paddle_whl: paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl |
| 183 | + run: | |
| 184 | + docker exec -t ${{ env.container_name }} /bin/bash -c ' |
| 185 | + echo "::group::Install bce-python-sdk" |
| 186 | + python -m pip install bce-python-sdk==0.8.74 |
| 187 | + echo "::endgroup::" |
| 188 | + export AK=paddle |
| 189 | + export SK=paddle |
| 190 | + if [ ! -f "${{ env.bos_file }}" ]; then |
| 191 | + 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 |
| 192 | + mkdir ${{ env.home_path }}/bos_retry |
| 193 | + tar xf ${{ env.home_path }}/bos_retry.tar.gz -C ${{ env.home_path }}/bos_retry |
| 194 | + fi |
| 195 | + cd /paddle |
| 196 | + mv /Paddle.tar.gz . |
| 197 | + cp ./build/python/dist/paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl . |
| 198 | + echo "Uploading Paddle.tar.gz" |
| 199 | + python ${{ env.bos_file }} Paddle.tar.gz paddle-github-action/PR/h-coverage/${{ env.PR_ID }}/${{ env.COMMIT_ID }} |
| 200 | + echo "Uploading coverage wheel" |
| 201 | + python ${{ env.bos_file }} ${{ env.paddle_whl }} paddle-github-action/PR/h-coverage/${{ env.PR_ID }}/${{ env.COMMIT_ID }} |
| 202 | + echo "End Upload" |
| 203 | + ' |
| 204 | +
|
| 205 | + - name: Terminate and delete the container |
| 206 | + if: ${{ steps.check-bypass.outputs.can-skip != 'true' && always() }} |
| 207 | + run: | |
| 208 | + set +e |
| 209 | + docker exec -t ${{ env.container_name }} /bin/bash -c 'rm -rf * .[^.]*' |
| 210 | + docker stop ${{ env.container_name }} |
| 211 | + docker rm ${{ env.container_name }} |
| 212 | +
|
| 213 | + test: |
| 214 | + name: Coverage test |
| 215 | + needs: [build] |
| 216 | + runs-on: |
| 217 | + group: H-Coverage |
| 218 | + steps: |
| 219 | + - name: Determine the runner |
| 220 | + run: | |
| 221 | + runner_name=`(echo $PWD|awk -F '/' '{print $3}')` |
| 222 | + echo $runner_name |
| 223 | + wget -q https://xly-devops.bj.bcebos.com/utils.sh |
| 224 | + source utils.sh |
| 225 | + determine_gpu_runner ${runner_name} |
| 226 | +
|
| 227 | + - name: Check docker image and run container |
| 228 | + env: |
| 229 | + CACHE_DIR: "/root/.cache/coverage" |
| 230 | + CCACHE_DIR: "/root/.ccache/coverage" |
| 231 | + FLAGS_fraction_of_gpu_memory_to_use: 0.15 |
| 232 | + CTEST_PARALLEL_LEVEL: 2 |
| 233 | + WITH_GPU: "ON" |
| 234 | + CUDA_ARCH_NAME: Hopper |
| 235 | + WITH_AVX: "ON" |
| 236 | + COVERALLS_UPLOAD: "ON" |
| 237 | + PADDLE_VERSION: 0.0.0 |
| 238 | + WITH_DISTRIBUTE: "ON" |
| 239 | + WITH_UNITY_BUILD: "ON" |
| 240 | + PY_VERSION: "3.10" |
| 241 | + WITH_SHARED_PHI: "ON" |
| 242 | + GPU_DEVICES: ${{ env.GPU_DEVICES }} |
| 243 | + WITH_CINN: "ON" |
| 244 | + INFERENCE_DEMO_INSTALL_DIR: /root/.cache/coverage |
| 245 | + CCACHE_MAXSIZE: 200G |
| 246 | + CCACHE_LIMIT_MULTIPLE: 0.8 |
| 247 | + FLAGS_PIR_OPTEST: "TRUE" |
| 248 | + ON_INFER: "ON" |
| 249 | + COVERAGE_FILE: ${{ github.workspace }}/build/python-coverage.data |
| 250 | + GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} |
| 251 | + GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }} |
| 252 | + run: | |
| 253 | + container_name=${TASK}-$(date +%Y%m%d-%H%M%S) |
| 254 | + echo "container_name=${container_name}" >> ${{ github.env }} |
| 255 | + docker_image=ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:cuda129-coverage-test |
| 256 | + docker run -d -t --gpus "\"device=${GPU_DEVICES}\"" --name ${container_name} \ |
| 257 | + -v "/home/data/cfs:/home/data/cfs" \ |
| 258 | + -v "/home/data/cfs/.cache:/root/.cache" \ |
| 259 | + -v "/home/data/cfs/.ccache:/root/.ccache" \ |
| 260 | + -v "/dev/shm:/dev/shm" \ |
| 261 | + -v ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \ |
| 262 | + -v ${{ github.workspace }}:/paddle \ |
| 263 | + -e CI_name \ |
| 264 | + -e BRANCH \ |
| 265 | + -e PR_ID \ |
| 266 | + -e COMMIT_ID \ |
| 267 | + -e work_dir \ |
| 268 | + -e PADDLE_ROOT \ |
| 269 | + -e GIT_PR_ID \ |
| 270 | + -e CACHE_DIR \ |
| 271 | + -e CCACHE_DIR \ |
| 272 | + -e ci_scripts \ |
| 273 | + -e FLAGS_fraction_of_gpu_memory_to_use \ |
| 274 | + -e CTEST_PARALLEL_LEVEL \ |
| 275 | + -e WITH_GPU \ |
| 276 | + -e CUDA_ARCH_NAME \ |
| 277 | + -e WITH_AVX \ |
| 278 | + -e WITH_COVERAGE \ |
| 279 | + -e COVERALLS_UPLOAD \ |
| 280 | + -e PADDLE_VERSION \ |
| 281 | + -e WITH_DISTRIBUTE \ |
| 282 | + -e WITH_UNITY_BUILD \ |
| 283 | + -e PY_VERSION \ |
| 284 | + -e WITH_SHARED_PHI \ |
| 285 | + -e WITH_CINN \ |
| 286 | + -e INFERENCE_DEMO_INSTALL_DIR \ |
| 287 | + -e CCACHE_MAXSIZE \ |
| 288 | + -e CCACHE_LIMIT_MULTIPLE \ |
| 289 | + -e FLAGS_PIR_OPTEST \ |
| 290 | + -e ON_INFER \ |
| 291 | + -e COVERAGE_FILE \ |
| 292 | + -e GITHUB_TOKEN \ |
| 293 | + -e GITHUB_API_TOKEN \ |
| 294 | + -e CFS_DIR \ |
| 295 | + -e no_proxy \ |
| 296 | + -w /paddle --network host ${docker_image} |
| 297 | +
|
| 298 | + - name: Download paddle.tar.gz and update test branch |
| 299 | + run: | |
| 300 | + docker exec -t ${{ env.container_name }} /bin/bash -c ' |
| 301 | + rm -rf * .[^.]* |
| 302 | + set -e |
| 303 | + echo "Downloading Paddle.tar.gz from cfs" |
| 304 | + wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PR/h-coverage/${PR_ID}/${COMMIT_ID}/Paddle.tar.gz --no-check-certificate |
| 305 | + echo "Extracting Paddle.tar.gz" |
| 306 | + tar --use-compress-program="pzstd -1" -xf Paddle.tar.gz --strip-components=1 |
| 307 | + rm Paddle.tar.gz |
| 308 | + ' |
| 309 | +
|
| 310 | + - name: Test |
| 311 | + run: | |
| 312 | + docker exec -t ${{ env.container_name }} /bin/bash -c ' |
| 313 | + source ${{ github.workspace }}/../../../proxy |
| 314 | + pip install build//python/dist/*.whl --no-deps |
| 315 | + pip install -r python/unittest_py/requirements.txt |
| 316 | + bash $ci_scripts/h-test.sh |
| 317 | + ' |
| 318 | +
|
| 319 | + - name: Terminate and delete the container |
| 320 | + if: always() |
| 321 | + run: | |
| 322 | + set +e |
| 323 | + rm Paddle.tar.gz |
| 324 | + docker exec -t ${{ env.container_name }} /bin/bash -c 'rm -rf * .[^.]*' |
| 325 | + docker stop ${{ env.container_name }} |
| 326 | + docker rm ${{ env.container_name }} |
0 commit comments