(undone)safe_clip #228
Workflow file for this run
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: CI Validation (MUSA GPU) | |
| on: | |
| pull_request: | |
| branches: [ main, develop ] | |
| types: [ opened, synchronize, reopened ] | |
| workflow_dispatch: | |
| inputs: | |
| target_branch: | |
| description: 'Target branch to check rebase against (for manual run)' | |
| required: false | |
| default: 'main' | |
| type: string | |
| schedule: | |
| - cron: '0 0 * * *' | |
| permissions: | |
| contents: read | |
| jobs: | |
| validate: | |
| runs-on: [self-hosted, musa-gpu] | |
| env: | |
| COMMIT_ID: ${{ github.event.pull_request.head.sha || github.sha }} | |
| LOG_BASE: /home/runner/ci_logs | |
| WORKSPACE: /home/runner/action-runner/_work/${{ github.event.repository.name }}/${{ github.event.repository.name }} | |
| steps: | |
| - name: Check who am I | |
| run: | | |
| echo "我是用户: $(whoami)" | |
| echo "我当前所在的家目录是: $HOME" | |
| # ============================== | |
| # Checkout | |
| # ============================== | |
| - name: Checkout code | |
| uses: actions/checkout@v4 | |
| with: | |
| fetch-depth: 0 | |
| ref: ${{ env.COMMIT_ID }} | |
| # ============================== | |
| # 创建日志目录 | |
| # ============================== | |
| - name: Prepare log directory | |
| run: | | |
| if [ "${{ github.event_name }}" == "pull_request" ]; then | |
| TYPE="pr" | |
| ID="${{ github.event.pull_request.head.sha }}" | |
| elif [ "${{ github.event_name }}" == "schedule" ]; then | |
| TYPE="daily" | |
| ID=$(date +%F) | |
| else | |
| TYPE="manual" | |
| ID="${{ github.run_id }}" | |
| fi | |
| LOG_DIR="$LOG_BASE/$TYPE/$ID" | |
| mkdir -p "$LOG_DIR" | |
| echo "LOG_DIR=$LOG_DIR" >> $GITHUB_ENV | |
| echo "TYPE=$TYPE" >> $GITHUB_ENV | |
| echo "✅ Log dir: $LOG_DIR" | |
| # ============================== | |
| # Rebase 检查（仅 PR） | |
| # ============================== | |
| # - name: Check Rebase Status | |
| # if: github.event_name == 'pull_request' | |
| # run: | | |
| # TARGET_BRANCH="${{ github.base_ref }}" | |
| # git fetch origin $TARGET_BRANCH | |
| # if ! git merge-base --is-ancestor origin/$TARGET_BRANCH HEAD; then | |
| # echo "❌ NOT rebased" | |
| # exit 1 | |
| # fi | |
| # ============================== | |
| # Format | |
| # ============================== | |
| - name: Check code formatting | |
| id: run-format | |
| run: | | |
| set +e | |
| exit_code=0 | |
| # 使用 bash -l -c 模拟登录，自动激活 Conda 环境 | |
| bash -l -c " | |
| set -o pipefail | |
| cd $WORKSPACE && | |
| LOG_FILE='$LOG_DIR/format.log' && | |
| FILE_LIST=\$(mktemp) && | |
| if [ '${{ github.event_name }}' = 'pull_request' ]; then | |
| TARGET_BRANCH='${{ github.base_ref }}' && | |
| git fetch origin \$TARGET_BRANCH && | |
| git diff -z --diff-filter=ACMRTUXB --name-only origin/\$TARGET_BRANCH...HEAD -- \ | |
| '*.c' '*.cc' '*.cpp' '*.cu' '*.h' '*.hpp' > \$FILE_LIST | |
| else | |
| find . \( -path ./build -o -path ./.git \) -prune -o \ | |
| -regex '.*\.\(cc\|cpp\|hpp\|c\|h\|cu\)' -print0 > \$FILE_LIST | |
| fi && | |
| if [ ! -s \$FILE_LIST ]; then | |
| echo 'No C/C++ files to format-check.' > \$LOG_FILE | |
| else | |
| xargs -0 clang-format --Werror --dry-run < \$FILE_LIST > \$LOG_FILE 2>&1 | |
| fi | |
| " || exit_code=$? | |
| if [ $exit_code -eq 0 ]; then | |
| echo "status=success" >> $GITHUB_OUTPUT | |
| else | |
| echo "status=failure" >> $GITHUB_OUTPUT | |
| exit 1 | |
| fi | |
| continue-on-error: true | |
| # ============================== | |
| # Build | |
| # ============================== | |
| - name: Build | |
| id: run-build | |
| run: | | |
| set +e | |
| exit_code=0 | |
| bash -l -c " | |
| export CPLUS_INCLUDE_PATH=/usr/include/c++/11:/usr/include/x86_64-linux-gnu/c++/11:\$CPLUS_INCLUDE_PATH && | |
| cd $WORKSPACE && | |
| cp -r /home/runner/tensorflow_musa_extension/CMakeLists.txt ./ && | |
| LOG_FILE='$LOG_DIR/build.log' && | |
| rm -rf ./build && | |
| ./build.sh > \$LOG_FILE 2>&1 | |
| " || exit_code=$? | |
| if [ $exit_code -eq 0 ]; then | |
| echo "status=success" >> $GITHUB_OUTPUT | |
| else | |
| echo "status=failure" >> $GITHUB_OUTPUT | |
| exit 1 | |
| fi | |
| continue-on-error: true | |
| # ============================== | |
| # Test | |
| # ============================== | |
| - name: Integration Test | |
| id: run-integration | |
| run: | | |
| set +e | |
| exit_code=0 | |
| bash -l -c " | |
| cd $WORKSPACE/test && | |
| LOG_FILE='$LOG_DIR/test.log' && | |
| timeout 30m python test_runner.py --fusion > \$LOG_FILE 2>&1 | |
| " || exit_code=$? | |
| if [ $exit_code -eq 0 ]; then | |
| echo "status=success" >> $GITHUB_OUTPUT | |
| else | |
| echo "status=failure" >> $GITHUB_OUTPUT | |
| exit 1 | |
| fi | |
| continue-on-error: true | |
| # ============================== | |
| # 测试整网性能 part1 | |
| # ============================== | |
| - name: Check Performance part1 | |
| id: run-performance-part1 | |
| run: | | |
| set -euo pipefail | |
| exit_code=0 | |
| bash -l -c " | |
| cd /home/runner/tf_test_model/prunedGraph && | |
| LOG_FILE='$LOG_DIR/performance-part1.log' && | |
| timeout 20m python run_graph_tf_musa.py --inference-only --inference-rounds 1000 > \$LOG_FILE 2>&1 | |
| " || exit_code=$? | |
| if [ $exit_code -ne 0 ]; then | |
| echo "status=failure" >> $GITHUB_OUTPUT | |
| exit 1 | |
| fi | |
| CURRENT_TIME=$(grep "Average Time (s)" "$LOG_DIR/performance-part1.log" | tail -1 | awk -F'|' '{print $3}' | tr -d ' ') | |
| if [ -z "$CURRENT_TIME" ]; then | |
| echo "❌ 无法解析性能结果" | |
| echo "status=failure" >> $GITHUB_OUTPUT | |
| exit 1 | |
| fi | |
| BASELINE=0.038332 | |
| THRESHOLD=$(awk -v base="$BASELINE" 'BEGIN {printf "%.6f", base * 1.05}') | |
| echo "Baseline : $BASELINE" | |
| echo "Threshold: $THRESHOLD" | |
| echo "Current : $CURRENT_TIME" | |
| if awk -v cur="$CURRENT_TIME" -v thr="$THRESHOLD" 'BEGIN {exit !(cur > thr)}'; then | |
| echo "❌ 性能退化" | |
| echo "status=failure" >> $GITHUB_OUTPUT | |
| exit 1 | |
| else | |
| echo "✅ 性能正常" | |
| echo "status=success" >> $GITHUB_OUTPUT | |
| fi | |
| continue-on-error: true | |
| # ============================== | |
| # 测试整网性能 part2 | |
| # ============================== | |
| - name: Check Performance part2 | |
| id: run-performance-part2 | |
| run: | | |
| set -euo pipefail | |
| exit_code=0 | |
| bash -l -c " | |
| cd /home/runner/tf_test_model/prunedGraph && | |
| LOG_FILE='$LOG_DIR/performance-part2.log' && | |
| timeout 20m python run_graph_tf_musa.py --compare-accuracy --rtol 1e-2 --atol 1e-2 > \$LOG_FILE 2>&1 | |
| " || exit_code=$? | |
| if [ $exit_code -ne 0 ]; then | |
| echo "❌ 精度测试脚本执行失败" | |
| echo "status=failure" >> $GITHUB_OUTPUT | |
| exit 1 | |
| fi | |
| if grep -q "精度对比通过！CPU 与 MUSA 结果一致" "$LOG_DIR/performance-part2.log"; then | |
| echo "✅ 精度对比通过" | |
| echo "status=success" >> $GITHUB_OUTPUT | |
| else | |
| echo "❌ 精度对比失败" | |
| echo "status=failure" >> $GITHUB_OUTPUT | |
| echo "===== 日志最后50行 =====" | |
| tail -50 "$LOG_DIR/performance-part2.log" | |
| exit 1 | |
| fi | |
| continue-on-error: true | |
| # ============================== | |
| # Summary（写入日志目录） | |
| # ============================== | |
| - name: Save Summary | |
| if: always() | |
| run: | | |
| SUMMARY="$LOG_DIR/summary.md" | |
| echo '# CI Report - $COMMIT_ID' > "$SUMMARY" | |
| echo "- Format: ${{ steps.run-format.outputs.status || 'skipped' }}" >> "$SUMMARY" | |
| echo "- Build: ${{ steps.run-build.outputs.status || 'skipped' }}" >> "$SUMMARY" | |
| echo "- Test: ${{ steps.run-integration.outputs.status || 'skipped' }}" >> "$SUMMARY" | |
| echo "- Performance: ${{ steps.run-performance-part1.outputs.status || 'skipped' }}" >> "$SUMMARY" | |
| echo "- Performance: ${{ steps.run-performance-part2.outputs.status || 'skipped' }}" >> "$SUMMARY" | |
| # ============================== | |
| # CI Summary（GitHub UI） | |
| # ============================== | |
| - name: CI Summary | |
| if: always() | |
| run: | | |
| echo "## CI Result" >> $GITHUB_STEP_SUMMARY | |
| echo "Commit: $COMMIT_ID" >> $GITHUB_STEP_SUMMARY | |
| echo "" >> $GITHUB_STEP_SUMMARY | |
| echo "| Step | Result |" >> $GITHUB_STEP_SUMMARY | |
| echo "|------|--------|" >> $GITHUB_STEP_SUMMARY | |
| echo "| Format | ${{ steps.run-format.outputs.status || 'skipped' }} |" >> $GITHUB_STEP_SUMMARY | |
| echo "| Build | ${{ steps.run-build.outputs.status || 'skipped' }} |" >> $GITHUB_STEP_SUMMARY | |
| echo "| Test | ${{ steps.run-integration.outputs.status || 'skipped' }} |" >> $GITHUB_STEP_SUMMARY | |
| echo "| Performance | ${{ steps.run-performance-part1.outputs.status || 'skipped' }} |" >> $GITHUB_STEP_SUMMARY | |
| echo "| Performance | ${{ steps.run-performance-part2.outputs.status || 'skipped' }} |" >> $GITHUB_STEP_SUMMARY | |
| if [ "${{ steps.run-build.outputs.status }}" == "failure" ] || \ | |
| [ "${{ steps.run-integration.outputs.status }}" == "failure" ] || \ | |
| [ "${{ steps.run-performance-part1.outputs.status }}" == "failure" ] || \ | |
| [ "${{ steps.run-performance-part2.outputs.status }}" == "failure" ]; then | |
| exit 1 | |
| fi |