|
| 1 | +name: Linux-XPU |
| 2 | + |
| 3 | +on: |
| 4 | + workflow_call: |
| 5 | + |
| 6 | +env: |
| 7 | + dockerfile: dockerfile |
| 8 | + docker_image: aa13dc110ab3 |
| 9 | + PR_ID: ${{ github.event.pull_request.number }} |
| 10 | + COMMIT_ID: ${{ github.event.pull_request.head.sha }} |
| 11 | + ci_scripts: /paddle/ci |
| 12 | + ci_scripts_runner: ${{ github.workspace }}/ci |
| 13 | + work_dir: /paddle |
| 14 | + PADDLE_ROOT: /paddle |
| 15 | + BRANCH: ${{ github.event.pull_request.base.ref }} |
| 16 | + CI_name: xpu |
| 17 | + |
| 18 | +defaults: |
| 19 | + run: |
| 20 | + shell: bash |
| 21 | + |
| 22 | +jobs: |
| 23 | + check-bypass: |
| 24 | + name: Check bypass for XPU |
| 25 | + uses: ./.github/workflows/check-bypass.yml |
| 26 | + with: |
| 27 | + workflow-name: 'xpu' |
| 28 | + secrets: |
| 29 | + github-token: ${{ secrets.GITHUB_TOKEN }} |
| 30 | + |
| 31 | + build: |
| 32 | + name: Build |
| 33 | + needs: check-bypass |
| 34 | + if: ${{ github.repository_owner == 'PaddlePaddle' && needs.check-bypass.outputs.can-skip != 'true' }} |
| 35 | + env: |
| 36 | + TASK: paddle-CI-${{ github.event.pull_request.number }}-xpu_build |
| 37 | + runs-on: |
| 38 | + group: Kunlun-CPU |
| 39 | + |
| 40 | + steps: |
| 41 | + - name: Download paddle.tar.gz and update test branch |
| 42 | + run: | |
| 43 | + set -e |
| 44 | + echo "Downloading Paddle.tar.gz" |
| 45 | + wget -q --no-proxy https://paddle-github-action.bj.bcebos.com/PR/Paddle/${PR_ID}/${COMMIT_ID}/Paddle.tar.gz --no-check-certificate |
| 46 | + echo "Extracting Paddle.tar.gz" |
| 47 | + tar xf Paddle.tar.gz --strip-components=1 |
| 48 | + rm Paddle.tar.gz |
| 49 | + git config --global user.name "PaddleCI" |
| 50 | + git config --global user.email "paddle_ci@example.com" |
| 51 | + git remote add upstream https://github.com/PaddlePaddle/Paddle.git |
| 52 | + source ${{ github.workspace }}/../../../proxy |
| 53 | + git config pull.rebase false |
| 54 | + git checkout test |
| 55 | + echo "Pull upstream develop or target branch" |
| 56 | + git pull upstream $BRANCH --no-edit |
| 57 | +
|
| 58 | + - name: Check docker image and run container |
| 59 | + env: |
| 60 | + WITH_SHARED_PHI: "ON" |
| 61 | + WITH_XPU: "ON" |
| 62 | + COVERALLS_UPLOAD: "OFF" |
| 63 | + WITH_AVX: "OFF" |
| 64 | + GIT_PR_ID: ${{ github.event.pull_request.number }} |
| 65 | + PADDLE_VERSION: 0.0.0 |
| 66 | + WITH_TESTING: "ON" |
| 67 | + WITH_DISTRIBUTE: "ON" |
| 68 | + PY_VERSION: "3.10" |
| 69 | + XPU_VISIBLE_DEVICES: "0,1" |
| 70 | + CUDA_VERSION: |
| 71 | + CUDNN_VERSION: |
| 72 | + WITH_XPU_BKCL: "ON" |
| 73 | + WITH_XPU_XRE5: "ON" |
| 74 | + CACHE_DIR: /root/.cache |
| 75 | + CCACHE_DIR: /root/.ccache |
| 76 | + CCACHE_MAXSIZE: 150G |
| 77 | + CCACHE_LIMIT_MULTIPLE: 0.8 |
| 78 | + no_proxy: "bcebos.com,apiin.im.baidu.com,gitee.com,aliyun.com,.baidu.com,.tuna.tsinghua.edu.cn" |
| 79 | + GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }} |
| 80 | + home_dir: ${{ github.workspace }}/../../../.. |
| 81 | + run: | |
| 82 | + container_name=${TASK}-$(date +%Y%m%d-%H%M%S) |
| 83 | + echo "container_name=${container_name}" >> ${{ github.env }} |
| 84 | + docker run --privileged --ulimit nofile=102400:102400 -d -t --name ${container_name} \ |
| 85 | + -v $home_dir/.cache:/root/.cache \ |
| 86 | + -v $home_dir/.ccache:/root/.ccache \ |
| 87 | + -v ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \ |
| 88 | + -v ${{ github.workspace }}:/paddle \ |
| 89 | + -e BRANCH \ |
| 90 | + -e PR_ID \ |
| 91 | + -e COMMIT_ID \ |
| 92 | + -e work_dir \ |
| 93 | + -e PADDLE_ROOT \ |
| 94 | + -e WITH_SHARED_PHI \ |
| 95 | + -e WITH_XPU \ |
| 96 | + -e COVERALLS_UPLOAD \ |
| 97 | + -e GIT_PR_ID \ |
| 98 | + -e PADDLE_VERSION \ |
| 99 | + -e WITH_TESTING \ |
| 100 | + -e WITH_DISTRIBUTE \ |
| 101 | + -e PY_VERSION \ |
| 102 | + -e XPU_VISIBLE_DEVICES \ |
| 103 | + -e CUDA_VERSION \ |
| 104 | + -e CUDNN_VERSION \ |
| 105 | + -e WITH_XPU_BKCL \ |
| 106 | + -e WITH_XPU_XRE5 \ |
| 107 | + -e WITH_INFERENCE_API_TEST \ |
| 108 | + -e CACHE_DIR \ |
| 109 | + -e CCACHE_DIR \ |
| 110 | + -e CCACHE_MAXSIZE \ |
| 111 | + -e CCACHE_LIMIT_MULTIPLE \ |
| 112 | + -e ci_scripts \ |
| 113 | + -e WITH_AVX \ |
| 114 | + -e no_proxy \ |
| 115 | + -e GITHUB_API_TOKEN \ |
| 116 | + -w /paddle --network host ${docker_image} /bin/bash |
| 117 | +
|
| 118 | + - name: Run build |
| 119 | + env: |
| 120 | + work_dir: ${{ github.workspace }} |
| 121 | + PADDLE_ROOT: ${{ github.workspace }} |
| 122 | + run: | |
| 123 | + docker exec -t ${{ env.container_name }} /bin/bash -c ' |
| 124 | + source ${{ github.workspace }}/../../../proxy |
| 125 | + ulimit -n 102400 |
| 126 | + git config --global --add safe.directory ${work_dir} |
| 127 | + git submodule foreach "git config --global --add safe.directory \$toplevel/\$sm_path" |
| 128 | + bash -x ${ci_scripts}/run_setup.sh bdist_wheel |
| 129 | + EXCODE=$? |
| 130 | + exit $EXCODE |
| 131 | + ' |
| 132 | +
|
| 133 | + - name: Upload build.tar.gz and paddle_whl to bos |
| 134 | + env: |
| 135 | + AK: paddle |
| 136 | + SK: paddle |
| 137 | + home_path: ${{ github.workspace }}/.. |
| 138 | + bos_file: ${{ github.workspace }}/../bos/BosClient.py |
| 139 | + paddle_whl: paddlepaddle_xpu-0.0.0-cp310-cp310-linux_x86_64.whl |
| 140 | + run: | |
| 141 | + if [ ! -f "${{ env.bos_file }}" ]; then |
| 142 | + wget -q --no-proxy -O ${{ env.home_path }}/bos_new.tar.gz https://xly-devops.bj.bcebos.com/home/bos_new.tar.gz --no-check-certificate |
| 143 | + mkdir ${{ env.home_path }}/bos |
| 144 | + tar xf ${{ env.home_path }}/bos_new.tar.gz -C ${{ env.home_path }}/bos |
| 145 | + fi |
| 146 | + cd .. |
| 147 | + tar --use-compress-program="pigz" -cpf build.tar.gz Paddle |
| 148 | + # source /home/opt/deck/1.0/etc/bashrc |
| 149 | + python3 ${bos_file} build.tar.gz paddle-github-action/PR/xpu/${{ env.PR_ID }}/${{ env.COMMIT_ID }} |
| 150 | + rm build.tar.gz |
| 151 | + cp ${{ github.workspace }}/dist/$paddle_whl . |
| 152 | + python3 ${bos_file} ${paddle_whl} paddle-github-action/PR/xpu/${{ env.PR_ID }}/${{ env.COMMIT_ID }} |
| 153 | + rm ${paddle_whl} |
| 154 | +
|
| 155 | + - name: Terminate and delete the container |
| 156 | + if: always() |
| 157 | + run: | |
| 158 | + docker exec -t ${container_name} /bin/bash -c 'rm -rf * .[^.]*' |
| 159 | + docker stop ${container_name} |
| 160 | + docker rm ${container_name} |
| 161 | +
|
| 162 | + test: |
| 163 | + name: Test |
| 164 | + needs: build |
| 165 | + env: |
| 166 | + TASK: paddle-CI-${{ github.event.pull_request.number }}-xpu_test |
| 167 | + runs-on: |
| 168 | + group: Kunlun |
| 169 | + |
| 170 | + steps: |
| 171 | + |
| 172 | + - name: Download build.tar.gz |
| 173 | + run: | |
| 174 | + sudo rm -rf * .[^.]* |
| 175 | + wget -q --no-proxy https://paddle-github-action.bj.bcebos.com/PR/xpu/${PR_ID}/${COMMIT_ID}/build.tar.gz --no-check-certificate |
| 176 | + # wget -q --no-proxy https://paddle-github-action.bj.bcebos.com/PR/xpu/72068/a58543d2f92bbe5c817e185e9cc24b549c919601/build.tar.gz --no-check-certificate |
| 177 | + tar --use-compress-program="pigz" -xpf build.tar.gz --strip-components=1 |
| 178 | + rm build.tar.gz |
| 179 | +
|
| 180 | + - name: Determine the runner |
| 181 | + run: | |
| 182 | + runner_name=`(echo $PWD|awk -F '/' '{print $3}')` |
| 183 | + echo $runner_name |
| 184 | + source ${ci_scripts_runner}/utils.sh |
| 185 | + determine_kunlun_runner ${runner_name} |
| 186 | +
|
| 187 | + - name: Check docker image and run container |
| 188 | + env: |
| 189 | + WITH_XPU: "ON" |
| 190 | + COVERALLS_UPLOAD: "OFF" |
| 191 | + CMAKE_BUILD_TYPE: Release |
| 192 | + WITH_AVX: "OFF" |
| 193 | + GIT_PR_ID: ${{ github.event.pull_request.number }} |
| 194 | + PADDLE_VERSION: 0.0.0 |
| 195 | + WITH_TESTING: "ON" |
| 196 | + WITH_DISTRIBUTE: "ON" |
| 197 | + PY_VERSION: "3.10" |
| 198 | + XPU_VISIBLE_DEVICES: ${{ env.CUDA_VISIBLE_DEVICES }} |
| 199 | + CUDA_VISIBLE_DEVICES: ${{ env.CUDA_VISIBLE_DEVICES }} |
| 200 | + CUDA_VERSION: |
| 201 | + CUDNN_VERSION: |
| 202 | + WITH_XPU_BKCL: "ON" |
| 203 | + CACHE_DIR: /root/.cache |
| 204 | + CCACHE_DIR: /root/.ccache |
| 205 | + CCACHE_MAXSIZE: 150G |
| 206 | + CCACHE_LIMIT_MULTIPLE: 0.8 |
| 207 | + no_proxy: "bcebos.com,apiin.im.baidu.com,gitee.com,aliyun.com,.baidu.com,.tuna.tsinghua.edu.cn" |
| 208 | + GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }} |
| 209 | + home_dir: ${{ github.workspace }}/../../../.. |
| 210 | + FLAGS_use_stride_kernel: "0" |
| 211 | + run: | |
| 212 | + container_name=${TASK}-$(date +%Y%m%d-%H%M%S) |
| 213 | + echo "container_name=${container_name}" >> ${{ github.env }} |
| 214 | + sudo mkdir -p /run/user/0 |
| 215 | + docker run --cap-add=SYS_PTRACE --privileged --ulimit nofile=102400 --ulimit core=-1 --shm-size=32g -d -t --name ${container_name} \ |
| 216 | + -v /ssd1/cibuild/.cache:/root/.cache \ |
| 217 | + -v /ssd1/cibuild/.ccache:/root/.ccache \ |
| 218 | + -v ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \ |
| 219 | + -v ${{ github.workspace }}:/paddle \ |
| 220 | + --device ${XPU_CODE_1} \ |
| 221 | + --device ${XPU_CODE_2} \ |
| 222 | + --shm-size=32g \ |
| 223 | + -e BRANCH \ |
| 224 | + -e PR_ID \ |
| 225 | + -e COMMIT_ID \ |
| 226 | + -e work_dir \ |
| 227 | + -e PADDLE_ROOT \ |
| 228 | + -e WITH_XPU \ |
| 229 | + -e COVERALLS_UPLOAD \ |
| 230 | + -e CMAKE_BUILD_TYPE \ |
| 231 | + -e GIT_PR_ID \ |
| 232 | + -e PADDLE_VERSION \ |
| 233 | + -e WITH_TESTING \ |
| 234 | + -e WITH_DISTRIBUTE \ |
| 235 | + -e PY_VERSION \ |
| 236 | + -e XPU_VISIBLE_DEVICES \ |
| 237 | + -e CUDA_VISIBLE_DEVICES \ |
| 238 | + -e CUDA_VERSION \ |
| 239 | + -e CUDNN_VERSION \ |
| 240 | + -e WITH_XPU_BKCL \ |
| 241 | + -e CACHE_DIR \ |
| 242 | + -e CCACHE_DIR \ |
| 243 | + -e WITH_INFERENCE_API_TEST \ |
| 244 | + -e CCACHE_MAXSIZE \ |
| 245 | + -e CCACHE_LIMIT_MULTIPLE \ |
| 246 | + -e ci_scripts \ |
| 247 | + -e WITH_AVX \ |
| 248 | + -e no_proxy \ |
| 249 | + -e GITHUB_API_TOKEN \ |
| 250 | + -e FLAGS_use_stride_kernel \ |
| 251 | + -w /paddle --network host ${docker_image} /bin/bash |
| 252 | +
|
| 253 | + - name: Run test |
| 254 | + run: | |
| 255 | + sudo mkdir -p /run/user/0 |
| 256 | + docker exec -t ${{ env.container_name }} /bin/bash -c ' |
| 257 | + bash ${ci_scripts}/kunlun_test.sh |
| 258 | + ' |
| 259 | +
|
| 260 | + - name: Terminate and delete the container |
| 261 | + if: always() |
| 262 | + run: | |
| 263 | + sudo mkdir -p /run/user/0 |
| 264 | + docker exec -t ${container_name} /bin/bash -c 'rm -rf * .[^.]*' |
| 265 | + docker stop ${container_name} |
| 266 | + docker rm ${container_name} |
0 commit comments