_910B-ARM.yml

name: PR-CI-910B-ARM

on:
  workflow_call:
    inputs:
      workflow-name:
        type: string
        required: false
      clone_dir:
        type: string
        required: false
        default: 'PaddlecustomDevice'
      is_pr:
        type: string
        required: false
        default: 'true'

jobs:
  check-bypass:
    name: Check bypass
    if: ${{ inputs.can-skip != 'true' }}
    uses: ./.github/workflows/check-bypass.yml
    with:
      workflow-name: '910b-arm'
    secrets:
      github-token: ${{ secrets.GITHUB_TOKEN }}

  PR-CI-910B-ARM:
    name: PR-CI-910B-ARM
    needs: check-bypass
    if: ${{ needs.check-bypass.outputs.can-skip != 'true' }}
    runs-on:
      group: 910B-ARM
    env:
      PR_ID: ${{ github.event.pull_request.number }}
      COMMIT_ID: ${{ github.event.pull_request.head.sha }}
      BRANCH: develop
      TASK: paddle-CI-${{ github.event.pull_request.number }}-npu
      no_proxy: "bcebos.com,apiin.im.baidu.com,gitee.com,aliyun.com,.baidu.com,.tuna.tsinghua.edu.cn"

    steps:
      - name: Check docker image and run container
        env:
          GIT_PR_ID: ${{ github.event.pull_request.number || '0' }}
          GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          RUN_ID: ${{ github.run_id }}
          cache_dir: /root/.cache
          ccache_dir: /root/.ccache
          docker_image: registry.baidubce.com/device/paddle-npu:cann80RC1-ubuntu20-aarch64-gcc84-py39
        run: |
          if [[ "${{ runner.name }}" == "IXUCA-2" ]];then
              echo "IXUCA-2"
              export ASCEND_RT_VISIBLE_DEVICES=4,5,6,7
          elif [[ "${{ runner.name }}" == "IXUCA-3" ]];then
              echo "IXUCA-3"
              export ASCEND_RT_VISIBLE_DEVICES=8,9,10,11
          elif [[ "${{ runner.name }}" == "IXUCA-4" ]];then
              echo "IXUCA-4"
              export ASCEND_RT_VISIBLE_DEVICES=12,13,14,15
          else
              echo echo "IXUCA-1"
              export ASCEND_RT_VISIBLE_DEVICES=0,1,2,3
          fi
          echo ${ASCEND_RT_VISIBLE_DEVICES}

          container_name=${TASK}-$(date +%Y%m%d-%H%M%S)
          echo "container_name=${container_name}" >> ${{ github.env }}
          docker run --shm-size=128G \
            -d -t --name ${container_name} \
            --privileged --network=host \
            -e PR_ID \
            -e COMMIT_ID \
            -e BRANCH \
            -e no_proxy \
            -e USE_910B=1 \
            -e FLAGS_npu_storage_format=0 \
            -e ASCEND_RT_VISIBLE_DEVICES=${ASCEND_RT_VISIBLE_DEVICES} \
            -v ${cache_dir}:/root/.cache \
            -v ${ccache_dir}:/root/.ccache \
            -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 ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \
            -v ${{ github.workspace }}:/PaddleCustomDevice -w /PaddleCustomDevice \
            ${docker_image} /bin/bash

      - name: Download Code
        run: |
          docker exec -t ${{ env.container_name }} /bin/bash -c '
          rm -rf /PaddleCustomDevice/* /PaddleCustomDevice/.[^.]* 2>/dev/null || true
          source ${{ github.workspace }}/../../../proxy
          set -x
          wget -q --tries=5 --no-proxy https://paddle-github-action.bj.bcebos.com/PaddleCustomDevice/PR/${PR_ID}/${COMMIT_ID}/PaddleCustomDevice.tar.gz --no-check-certificate
          echo "Extracting PaddleCustomDevice.tar.gz"
          tar -xf PaddleCustomDevice.tar.gz --strip-components=1
          git config --global --add safe.directory "*"
          git remote add upstream https://github.com/PaddlePaddle/PaddleCustomDevice.git
          git merge ${BRANCH} --no-edit
          git --no-pager log --pretty=oneline -5'

      - name: RUN 910B-ARM
        run: |
          docker exec -t ${{ env.container_name }} /bin/bash -c '
          # !!!!! SKIP IF NO NPU CHANGE !!!!
          echo "=========== Checking PR Changes If NPU FULL CI Needed ==========="
          change_numbers=$(git --no-pager diff --name-only remotes/origin/${BRANCH} | wc -l)
          change_backend=$(git --no-pager diff --name-only remotes/origin/${BRANCH} | grep "backends/"| wc -l)
          change_npu_only=$(git --no-pager diff --name-only remotes/origin/${BRANCH} | grep "backends/npu"| wc -l)
          git --no-pager diff --name-only remotes/origin/${BRANCH}

          if [ $change_numbers -ne $change_backend ]; then
            echo "Common file changed, continue to run NPU FULL CI test ..."
          elif [ $change_npu_only -eq 0 ] ; then
            echo "NO NPU backend changes found, skip NPU FULL CI ...."
            exit 0
          fi

          echo "============ CANN Version ============="
          source ~/.bashrc
          echo "============ Install PaddlePaddle CPU ============="
          PATH=/usr/local/bin:${PATH}
          ln -sf $(which python3.9) /usr/local/bin/python
          ln -sf $(which pip3.9) /usr/local/bin/pip

          pip install -U numpy==1.26.4 safetensors
          python -m pip install --pre paddlepaddle -i https://www.paddlepaddle.org.cn/packages/nightly/cpu/
          python -c "import paddle; print(paddle.__version__)"
          python -c "import paddle; print(paddle.version.commit)"
          # Install python packages required for incremental coverage
          python -m pip install PyGithub
          export LD_PRELOAD=/lib/aarch64-linux-gnu/libgomp.so.1:$LD_PRELOAD
          export PADDLE_VERSION=0.0.0
          bash backends/npu/tools/pr_ci_npu.sh'

      - name: Terminate and delete the container
        if: always()
        run: |
          docker exec -t ${{ env.container_name }} /bin/bash -c '
          rm -rf * .[^.]*
          chmod -R 777 . || true
          '
          docker stop ${container_name}
          docker rm ${container_name}