[iluvatar] iluvatar CI moves to k8s

.github/workflows/_IXUCA.yml

@@ -16,11 +16,25 @@ on:
         default: 'true'
 
 jobs:
-  PR-CI-HPU:
+  check-bypass:
+    name: Check bypass
+    if: ${{ inputs.can-skip != 'true' }}
+    uses: ./.github/workflows/check-bypass.yml
+    with:
+      workflow-name: 'iluvatar'
+    secrets:
+      github-token: ${{ secrets.GITHUB_TOKEN }}
+
+  PR-CI-IXUCA:
     name: PR-CI-IXUCA
-    if: ${{ github.repository_owner == 'PaddlePaddle' }}
-    runs-on:
-      group: test2
+    needs: check-bypass
+    if:  ${{ needs.check-bypass.outputs.can-skip != 'true' && github.repository_owner == 'PaddlePaddle' }}
+    runs-on: iluvatar-gpu-2
+    container:
+      image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:latest
+      env:
+        LD_LIBRARY_PATH: /usr/local/corex/lib
+        LIBRARY_PATH: /usr/local/corex/lib
     env:
       PR_ID: ${{ github.event.pull_request.number }}
       COMMIT_ID: ${{ github.event.pull_request.head.sha }}
@@ -29,37 +43,8 @@ jobs:
       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: ${{ github.run_id }}
-          cache_dir: /root/.cache
-          ccache_dir: /root/.ccache
-          docker_image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:latest
-        run: |
-          container_name=${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 LD_LIBRARY_PATH=/usr/local/corex-4.3.0/lib \
-            -e LIBRARY_PATH=/usr/local/corex-4.3.0/lib \
-            -v ${cache_dir}:/root/.cache \
-            -v ${ccache_dir}:/root/.ccache \
-            -v ${{ github.workspace }}/../../..:${{ github.workspace }}/../../.. \
-            -v ${{ github.workspace }}:/workspace -w /workspace \
-            ${docker_image} /bin/bash
-
       - name: Download Code
         run: |
-          docker exec -t ${{ env.container_name }} /bin/bash -c '
-          rm -rf /workspace/* /workspace/.[^.]* 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"
@@ -68,81 +53,64 @@ jobs:
           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: Check bypass
-        id: check-bypass
-        uses: ./PaddleCustomDevice/.github/actions/check-bypass
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-          workflow-name: ixuca
+          git --no-pager log --pretty=oneline -5
 
       - name: RUN IXUCA
         id: run-ixuca
         if: steps.check-bypass.outputs.can-skip != 'true'
         run: |
           echo "=========== Checking PR Changes If IXUCA FULL CI Needed ==========="
 
-          should_skip=$(docker exec -t ${{ env.container_name }} /bin/bash -c '
-            cd /workspace/PaddleCustomDevice
+          cd PaddleCustomDevice
 
-            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_ixuca_only=$(git --no-pager diff --name-only remotes/origin/${BRANCH} | grep "backends/iluvatar_gpu" | wc -l)
+          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_ixuca_only=$(git --no-pager diff --name-only remotes/origin/${BRANCH} | grep "backends/iluvatar_gpu" | wc -l)
 
-            echo "Changed files:"
-            git --no-pager diff --name-only remotes/origin/${BRANCH}
+          echo "Changed files:"
+          git --no-pager diff --name-only remotes/origin/${BRANCH}
 
-            if [ $change_numbers -ne $change_backend ]; then
-              echo "Common file changed, continue to run IXUCA FULL CI test ..."
-              echo false
-            elif [ $change_ixuca_only -eq 0 ]; then
-              echo "NO IXUCA backend changes found, skip IXUCA FULL CI ...."
-              echo true
-            else
-              echo "IXUCA backend changed, run FULL CI ..."
-              echo false
-            fi
-          ' | tail -n 1 | tr -d '\r')
+          should_skip="false"
+          if [ $change_numbers -ne $change_backend ]; then
+            echo "Common file changed, continue to run IXUCA FULL CI test ..."
+          elif [ $change_ixuca_only -eq 0 ]; then
+            echo "NO IXUCA backend changes found, skip IXUCA FULL CI ...."
+            should_skip="true"
+          else
+            echo "IXUCA backend changed, run FULL CI ..."
+          fi
 
-          echo "should_skip from docker = $should_skip"
           echo "should_skip=$should_skip" >> $GITHUB_OUTPUT
 
       - name: Build and Install paddle_iluvatar_gpu
         if: steps.check-bypass.outputs.can-skip != 'true' && steps.run-ixuca.outputs.should_skip != 'true'
         run: |
-          docker exec -t ${{ env.container_name }} /bin/bash -c '
-          export PATH=/usr/local/corex-4.3.0/bin:$PATH
-          mkdir -p /workspace/PaddleCustomDevice/Paddle/third_party/mklml/Linux
-          curl -o /workspace/PaddleCustomDevice/Paddle/third_party/mklml/Linux/csrmm_mklml_lnx_2019.0.5.tgz  http://paddlepaddledeps.bj.bcebos.com/csrmm_mklml_lnx_2019.0.5.tgz
-          tar xf /workspace/PaddleCustomDevice/Paddle/third_party/mklml/Linux/csrmm_mklml_lnx_2019.0.5.tgz
+          export PATH=/usr/local/corex/bin:$PATH
+          mkdir -p PaddleCustomDevice/Paddle/third_party/mklml/Linux
+          curl -o PaddleCustomDevice/Paddle/third_party/mklml/Linux/csrmm_mklml_lnx_2019.0.5.tgz  http://paddlepaddledeps.bj.bcebos.com/csrmm_mklml_lnx_2019.0.5.tgz
+          tar xf PaddleCustomDevice/Paddle/third_party/mklml/Linux/csrmm_mklml_lnx_2019.0.5.tgz
           python3 -m pip install --pre paddlepaddle -i https://www.paddlepaddle.org.cn/packages/nightly/cpu/
           pip show paddlepaddle
           python3 -m pip install parameterized
           export PADDLE_VERSION=0.0.0
-          cd /workspace/PaddleCustomDevice/backends/iluvatar_gpu
+          cd PaddleCustomDevice/backends/iluvatar_gpu
           # build
           bash build_paddle.sh
           # Install
-          bash install_paddle.sh'
+          bash install_paddle.sh
 
       - name: Run Tests
         if: steps.check-bypass.outputs.can-skip != 'true' && steps.run-ixuca.outputs.should_skip != 'true'
         run: |
-          docker exec -t ${{ env.container_name }} /bin/bash -c '
           pip show parameterized
           python3 -m pip install parameterized
           export FLAG_SKIP_FLOAT64=1
-          cd /workspace/PaddleCustomDevice/backends/iluvatar_gpu/tests
+          cd PaddleCustomDevice/backends/iluvatar_gpu/tests
           # Run tests
-          bash run_test.sh'
+          bash run_test.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}