This repository subset packages the code paths behind the public jaguar ReID ArXiv core claims and key experiment families.
- Axis-1 and Axis-2 claimset pipeline (manifest build, per-cell execution, aggregation).
- Claim gating and provenance tooling (paired AP/CMC significance, go/no-go checks, provenance and SoT helpers).
- Key benchmark families used in the paper narrative: ArcFace, Lorentz, and Triplet.
- ArXiv table traceability assets and deterministic CSV to TeX row export adapter.
- Optional HPI SLURM templates under
remote_setup/templates/with placeholder paths.
- Heavy datasets, checkpoints, W&B exports, and large canonical result bundles.
- Exploratory and legacy experiment branches not required by the core claims path.
python3 -m venv .venv
source .venv/bin/activate
pip install -e .Smoke and contract checks:
python3 tools/verify_release_bundle.py --bundle-root .Tiny local claimset flow (mocked inputs):
python3 tools/run_tiny_claimset_smoke.py --bundle-root . --workdir ./tmp_smokeArXiv table row export from canonical CSV artifacts:
python3 tools/export_arxiv_table_rows.py \
--canonical_root /path/to/canonical_axis12_test_bundle \
--output_dir ./generated_tablesCLAIM_TO_CODE_MATRIX.csvCLAIM_TO_CODE_MATRIX.mdREPRODUCIBILITY.mdEXPERIMENT_SCOPE.md