Infra: launch frozen staged V121 semantic residual #3
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: Trace Ace V121 Frozen Semantic PR Launch | |
| on: | |
| pull_request: | |
| branches: [main] | |
| workflow_dispatch: | |
| concurrency: | |
| group: trace-ace-v121-frozen-pr | |
| cancel-in-progress: true | |
| env: | |
| TRANSCRIPT_KEY: trace-ace-transcripts-v1-603547640 | |
| TRANSCRIPT_SHA: e685b85b04694e130c25b17d09cdd1892fbda5e9fa685e98b2300114b915aa2d | |
| SCIENCE_REF: agent/v145-v135-plus-evidence-family | |
| jobs: | |
| prepare: | |
| runs-on: ubuntu-24.04 | |
| timeout-minutes: 30 | |
| steps: | |
| - uses: actions/checkout@v4 | |
| with: | |
| ref: ${{ env.SCIENCE_REF }} | |
| - uses: actions/setup-python@v5 | |
| with: | |
| python-version: '3.12' | |
| cache: pip | |
| - name: Install dependencies | |
| run: python -m pip install --disable-pip-version-check numpy pandas scipy scikit-learn gdown fastembed==0.8.0 | |
| - name: Restore frozen transcript archive | |
| uses: actions/cache/restore@v4 | |
| with: | |
| path: transcripts.zip | |
| key: ${{ env.TRANSCRIPT_KEY }} | |
| fail-on-cache-miss: true | |
| - name: Validate frozen transcript archive | |
| run: | | |
| set -euo pipefail | |
| test "$(stat -c%s transcripts.zip)" = '603547640' | |
| test "$(sha256sum transcripts.zip | cut -d' ' -f1)" = "$TRANSCRIPT_SHA" | |
| - name: Download frozen metadata and extract | |
| run: | | |
| set -euo pipefail | |
| gdown 1EpqoamY0vFI2qE57R6wdqU5HwuoVk3Zz -O metadata.zip | |
| mkdir -p data/meta data/transcripts | |
| unzip -q metadata.zip -d data/meta | |
| unzip -q transcripts.zip -d data/transcripts | |
| echo "FEATURES=$(find data/meta -type f -name 'train_features*.csv' -print -quit)" >> "$GITHUB_ENV" | |
| echo "LABELS=$(find data/meta -type f -name 'train_labels*.csv' -print -quit)" >> "$GITHUB_ENV" | |
| FIRST=$(find data/transcripts -type f -name '*.csv' -print -quit) | |
| echo "TRANSCRIPTS=$(dirname "$FIRST")" >> "$GITHUB_ENV" | |
| - name: Prepare exact frozen V121 inputs | |
| run: | | |
| python -m py_compile competitions/trace_the_ace/v121_pretrained_semantic_residual.py competitions/trace_the_ace/v121_staged_transport.py | |
| cd competitions/trace_the_ace | |
| python v121_staged_transport.py prepare \ | |
| --features "../../$FEATURES" --labels "../../$LABELS" \ | |
| --transcripts "../../$TRANSCRIPTS" --rows 2500 --dir ../../v121_prepared | |
| - uses: actions/upload-artifact@v4 | |
| with: | |
| name: v121-prepared-pr | |
| path: v121_prepared/ | |
| retention-days: 2 | |
| compression-level: 6 | |
| embed: | |
| needs: prepare | |
| runs-on: ubuntu-24.04 | |
| timeout-minutes: 120 | |
| steps: | |
| - uses: actions/checkout@v4 | |
| with: | |
| ref: ${{ env.SCIENCE_REF }} | |
| - uses: actions/setup-python@v5 | |
| with: | |
| python-version: '3.12' | |
| cache: pip | |
| - name: Install dependencies | |
| run: python -m pip install --disable-pip-version-check numpy pandas scipy scikit-learn fastembed==0.8.0 | |
| - uses: actions/download-artifact@v4 | |
| with: | |
| name: v121-prepared-pr | |
| path: v121_prepared | |
| - name: Embed all 16 exact frozen shards sequentially | |
| run: | | |
| set -euo pipefail | |
| cd competitions/trace_the_ace | |
| for shard in $(seq 0 15); do | |
| echo "===== FROZEN V121 SHARD $shard / 15 =====" | |
| python v121_staged_transport.py embed --dir ../../v121_prepared \ | |
| --shard "$shard" --shards 16 \ | |
| --out "../../v121_embeddings_shard_${shard}.npz" | |
| done | |
| cd ../.. | |
| test "$(find . -maxdepth 1 -name 'v121_embeddings_shard_*.npz' | wc -l)" -eq 16 | |
| - uses: actions/upload-artifact@v4 | |
| with: | |
| name: v121-pr-embeddings-all | |
| path: v121_embeddings_shard_*.npz | |
| retention-days: 2 | |
| compression-level: 0 | |
| evaluate: | |
| needs: embed | |
| runs-on: ubuntu-24.04 | |
| timeout-minutes: 30 | |
| steps: | |
| - uses: actions/checkout@v4 | |
| with: | |
| ref: ${{ env.SCIENCE_REF }} | |
| - uses: actions/setup-python@v5 | |
| with: | |
| python-version: '3.12' | |
| cache: pip | |
| - name: Install dependencies | |
| run: python -m pip install --disable-pip-version-check numpy pandas scipy scikit-learn fastembed==0.8.0 | |
| - uses: actions/download-artifact@v4 | |
| with: | |
| name: v121-prepared-pr | |
| path: v121_prepared | |
| - uses: actions/download-artifact@v4 | |
| with: | |
| name: v121-pr-embeddings-all | |
| path: v121_embedding_shards | |
| - name: Evaluate frozen V121 precommit | |
| run: | | |
| cd competitions/trace_the_ace | |
| python v121_staged_transport.py evaluate --dir ../../v121_prepared \ | |
| --embeddings ../../v121_embedding_shards --out ../../v121_pretrained_semantic_residual.json | |
| - name: Show decision | |
| run: cat v121_pretrained_semantic_residual.json | |
| - uses: actions/upload-artifact@v4 | |
| with: | |
| name: trace-ace-v121-pretrained-semantic-residual-pr | |
| path: v121_pretrained_semantic_residual.json | |
| retention-days: 14 |