1# Reproducibility — WP41 CHRONOS-X23## What is needed4- A FinData OS API key with scopes `read:market, query, export` (private FirstRate Data license; the owner can issue one5 with `findata create-key`). Without it, the synthetic example below reproduces the software path only.6- Python 3.12, `uv`; an Apple Silicon Mac (MPS) or a CUDA/CPU machine (slower). Measured budgets: `docs/PHASE_4_REPORT.md`.78## Environment9```bash10make install # uv venv + editable install (torch, polars, pyarrow, lightgbm, arch, ...)11make test # 26 offline tests (+ 1 slow synthetic pipeline test with -m slow)12WP41_LIVE=1 make test-live # 5 live contract tests against FinData13```1415## License-free example (no FinData access)16```bash17.venv/bin/python scripts/make_synthetic.py # 4 synthetic instruments, FinData schema18.venv/bin/python scripts/run_experiment.py --config configs/synthetic_smoke.yaml19.venv/bin/python scripts/evaluate.py --exps synthetic_smoke # LaTeX tables from the results20```2122## Full study (in order; every step writes its artifacts under results/ and its provenance under results/logs/)23```bash24export FINDATA_API_KEY_FILE=~/findata-secrets/api-key-wp41.txt FINDATA_BASE_URL=https://findata.spboucher.ai/api/v125.venv/bin/python scripts/audit_api.py # P026.venv/bin/python scripts/build_universe.py # P1 universe v1 (PIT rules)27.venv/bin/python scripts/audit_coverage.py --candidates --workers 4 # P1 coverage audit28.venv/bin/python scripts/build_universe.py --min-coverage 0.8 \29 --coverage-file results/metrics/coverage_audit_candidates.csv --out universe_main_v2.json # -> data/manifests/universe_main.json30.venv/bin/python scripts/download_pilot.py --config configs/pilot.yaml --symbols-from pilot # P1 pilot bars + grids31.venv/bin/python scripts/run_experiment.py --config configs/pilot.yaml # P2/P3 pilot32.venv/bin/python scripts/benchmark.py --config configs/pilot.yaml # P4 profiling33.venv/bin/python scripts/download_pilot.py --config configs/main.yaml --symbols-from all # P4 main bars + grids34.venv/bin/python scripts/run_experiment.py --config configs/main.yaml # P5 main (dev test 2024)35.venv/bin/python scripts/run_experiment.py --config configs/ablations.yaml --skip-baselines # P5/P6 ablations + copula36.venv/bin/python scripts/run_experiment.py --config configs/main.yaml --heldout --skip-baselines # zero-shot instruments37.venv/bin/python scripts/run_experiment.py --config configs/final.yaml --eval-split final --models-from results/metrics/main # P7 untouched 202538.venv/bin/python scripts/evaluate.py --exps pilot,main,ablations,final # P8 tables + macros39cd paper && make # P9 PDF40```41Seeds: `train.seed` in each config (1; seeds 2 and 3 in `configs/ablations.yaml`); DataLoader generator seeded with42`seed + step`; evaluation sampling seeded. Checkpoints (`best.pt`, `last.pt`) resume automatically.4344## Artifact map45| Artifact | Path |46|---|---|47| Universe manifest (hash inside) | `data/manifests/universe_main.json` |48| API audit, coverage audit, download/grid stats | `results/metrics/p0_api_audit.json`, `coverage_audit_candidates.csv`, `download_*.json`, `grids_*.json` |49| Per-experiment outputs | `results/metrics/<exp>/` (summaries, per-origin scores, leaderboards, comparisons, train logs, checkpoints) |50| Benchmarks | `results/metrics/benchmark_<node>.json` |51| Provenance | `results/logs/provenance_*.jsonl`, `~/wp41-cache/manifests/bars_manifest.jsonl` (node) |52| Paper tables/macros | `paper/tables/*.tex`, `paper/results_macros.tex` (generated by `scripts/evaluate.py`) |53| Phase reports | `docs/PHASE_0..9_REPORT.md` |54