refactor(UCBD): delegate 11 duplicate primitives to trueno canonical … #48
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
| # Benchmark CI Workflow for whisper.apr | ||
|
Check failure on line 1 in .github/workflows/benchmark.yml
|
||
| # Reference: docs/specifications/benchmark-whisper-steps-a-z.md | ||
| # | ||
| # Triggers: | ||
| # - Manual (workflow_dispatch) | ||
| # - PR changes to src/**, benches/**, Cargo.toml | ||
| # - Weekly scheduled (Sundays 2 AM UTC) | ||
| name: Benchmarks | ||
| on: | ||
| workflow_dispatch: | ||
| inputs: | ||
| reason: | ||
| description: 'Reason for running benchmarks' | ||
| required: "false" | ||
| default: 'Manual benchmark run' | ||
| pull_request: | ||
| paths: | ||
| - 'src/**/*.rs' | ||
| - 'benches/**/*.rs' | ||
| - 'examples/**/*.rs' | ||
| - 'Cargo.toml' | ||
| - 'Cargo.lock' | ||
| schedule: | ||
| - cron: '0 2 * * 0' # Every Sunday at 2 AM UTC | ||
| env: | ||
| CARGO_TERM_COLOR: always | ||
| jobs: | ||
| benchmark: | ||
| name: Run Pipeline Benchmarks | ||
| runs-on: ubuntu-latest | ||
| steps: | ||
| - name: Checkout code | ||
| uses: actions/checkout@v4 | ||
| - name: Setup Rust toolchain | ||
| uses: dtolnay/rust-toolchain@stable | ||
| - name: Cache Rust dependencies | ||
| uses: Swatinem/rust-cache@v2 | ||
| - name: Download test models | ||
| run: | | ||
| mkdir -p models | ||
| # Check if models exist in cache, otherwise download | ||
| if [ ! -f models/whisper-tiny-int8.apr ]; then | ||
| echo "Note: Model files not in cache. Pipeline benchmarks will be skipped." | ||
| echo "To enable full benchmarks, add model files to the repository or cache." | ||
| fi | ||
| - name: Run preprocessing benchmarks (Steps B-F) | ||
| run: | | ||
| echo "=== Running Preprocessing Benchmarks ===" | ||
| cargo bench --bench pipeline -- step_b step_c step_f 2>&1 | tee bench-preprocessing.txt | ||
| continue-on-error: "true" | ||
| - name: Run format comparison benchmark | ||
| run: | | ||
| echo "=== Running Format Comparison ===" | ||
| cargo bench --bench format_comparison 2>&1 | tee bench-format.txt | ||
| continue-on-error: "true" | ||
| - name: Run format comparison example | ||
| run: | | ||
| echo "=== Running Format Comparison Example ===" | ||
| cargo run --release --example format_comparison 2>&1 | tee format-comparison-output.txt | ||
| continue-on-error: "true" | ||
| - name: Generate benchmark summary | ||
| if: always() | ||
| run: | | ||
| cat > benchmark-summary.md << 'EOF' | ||
| # Whisper.apr Benchmark Results | ||
| **Date:** $(date -u) | ||
| **Branch:** ${{ github.ref_name }} | ||
| **Commit:** ${{ github.sha }} | ||
| EOF | ||
| if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then | ||
| echo "**Trigger:** Manual (${{ github.event.inputs.reason }})" >> benchmark-summary.md | ||
| elif [ "${{ github.event_name }}" = "pull_request" ]; then | ||
| echo "**Trigger:** Pull Request #${{ github.event.pull_request.number }}" >> benchmark-summary.md | ||
| elif [ "${{ github.event_name }}" = "schedule" ]; then | ||
| echo "**Trigger:** Weekly Schedule" >> benchmark-summary.md | ||
| fi | ||
| cat >> benchmark-summary.md << 'EOF' | ||
| ## Pipeline Steps Benchmarked | ||
| | Step | Description | Status | | ||
| |------|-------------|--------| | ||
| | B | WAV file load | See results | | ||
| | C | PCM parse | See results | | ||
| | F | Mel spectrogram | See results | | ||
| | G | Encoder forward | Requires model | | ||
| | H | Decoder forward | Requires model | | ||
| ## Format Comparison | ||
| See `format-comparison-output.txt` for detailed results. | ||
| ## Performance Targets (WAPR-BENCH-001) | ||
| | Component | Target | Status | | ||
| |-----------|--------|--------| | ||
| | Mel (1.5s) | <10ms | ⏳ | | ||
| | Encoder | <500ms | ⏳ | | ||
| | Decoder/token | <100ms | ⏳ | | ||
| | First Token | <800ms | ⏳ | | ||
| | RTF (1.5s) | <2.0x | ⏳ | | ||
| ## Notes | ||
| - Full encoder/decoder benchmarks require model files | ||
| - Run `make golden-traces` locally for baseline capture | ||
| - Reference: `docs/specifications/benchmark-whisper-steps-a-z.md` | ||
| EOF | ||
| cat benchmark-summary.md | ||
| - name: Upload criterion results | ||
| if: always() | ||
| uses: actions/upload-artifact@v4 | ||
| with: | ||
| name: criterion-results-${{ github.sha }} | ||
| path: target/criterion/ | ||
| retention-days: 90 | ||
| - name: Upload benchmark outputs | ||
| if: always() | ||
| uses: actions/upload-artifact@v4 | ||
| with: | ||
| name: benchmark-output-${{ github.sha }} | ||
| path: | | ||
| bench-*.txt | ||
| format-comparison-output.txt | ||
| benchmark-summary.md | ||
| retention-days: 30 | ||
| - name: Comment on PR (if applicable) | ||
| if: github.event_name == 'pull_request' && always() | ||
| uses: actions/github-script@v7 | ||
| with: | ||
| script: | | ||
| const fs = require('fs'); | ||
| let summary = ''; | ||
| try { | ||
| summary = fs.readFileSync('benchmark-summary.md', 'utf8'); | ||
| } catch (e) { | ||
| summary = 'Benchmark summary not available.'; | ||
| } | ||
| let formatOutput = ''; | ||
| try { | ||
| formatOutput = fs.readFileSync('format-comparison-output.txt', 'utf8'); | ||
| // Truncate if too long | ||
| if (formatOutput.length > 3000) { | ||
| formatOutput = formatOutput.substring(0, 3000) + '\n... (truncated)'; | ||
| } | ||
| } catch (e) { | ||
| formatOutput = 'Format comparison output not available.'; | ||
| } | ||
| const body = `## 📊 Benchmark Results | ||
| ${summary} | ||
| <details> | ||
| <summary>Format Comparison Output</summary> | ||
| \`\`\` | ||
| ${formatOutput} | ||
| \`\`\` | ||
| </details> | ||
| --- | ||
| *Benchmarks completed on commit ${context.sha.substring(0, 7)}*`; | ||
| await github.rest.issues.createComment({ | ||
| issue_number: context.issue.number, | ||
| owner: context.repo.owner, | ||
| repo: context.repo.name, | ||
| body: body | ||
| }); | ||
