LOBench-A-share-processed / code /experiments /tbd_revision /scripts /run_label_threshold_retraining.py
| #!/usr/bin/env python3 | |
| """Retrain the five-stock prediction comparison under alternate labels.""" | |
| from __future__ import annotations | |
| import argparse | |
| import copy | |
| import json | |
| import sys | |
| from pathlib import Path | |
| ROOT = Path(__file__).resolve().parents[3] | |
| SCRIPT_DIR = ROOT / "experiments/tbd_revision/scripts" | |
| if str(SCRIPT_DIR) not in sys.path: | |
| sys.path.insert(0, str(SCRIPT_DIR)) | |
| from train_representation_baselines import run_one, summarize # noqa: E402 | |
| STOCKS = ("asset_001", "asset_002", "asset_017", "asset_005", "asset_033") | |
| METHODS = ("timae", "lobgrid_timae", "simlob2026") | |
| SEEDS = (2026, 2027, 2028) | |
| THRESHOLDS = { | |
| "threshold0001": 0.0001, | |
| "threshold0005": 0.0005, | |
| "threshold0010": 0.0010, | |
| } | |
| def parse_args() -> argparse.Namespace: | |
| parser = argparse.ArgumentParser(description=__doc__) | |
| parser.add_argument("--base-config", type=Path, required=True) | |
| parser.add_argument("--output-root", type=Path, required=True) | |
| parser.add_argument( | |
| "--variants", | |
| nargs="+", | |
| choices=tuple(THRESHOLDS), | |
| required=True, | |
| ) | |
| parser.add_argument("--force", action="store_true") | |
| return parser.parse_args() | |
| def resolved(path: Path) -> Path: | |
| return path if path.is_absolute() else ROOT / path | |
| def require_pretraining(config: dict) -> None: | |
| directory = resolved(Path(config["pretrain_dir"])) | |
| pretrain_id = str(config["pretrain_id"]) | |
| missing = [] | |
| for method in METHODS: | |
| for seed in SEEDS: | |
| stem = directory / f"{pretrain_id}_{method}_seed{seed}_encoder" | |
| for suffix in (".pt", ".pretrain.json"): | |
| path = stem.with_suffix(suffix) | |
| if not path.is_file(): | |
| missing.append(str(path)) | |
| if missing: | |
| raise FileNotFoundError( | |
| "The retained pooled checkpoints must be complete before label " | |
| f"retraining. Missing examples: {missing[:3]}" | |
| ) | |
| def main() -> None: | |
| args = parse_args() | |
| base = json.loads(resolved(args.base_config).read_text(encoding="utf-8")) | |
| require_pretraining(base) | |
| output_root = resolved(args.output_root) | |
| output_root.mkdir(parents=True, exist_ok=True) | |
| for variant in args.variants: | |
| config = copy.deepcopy(base) | |
| threshold = THRESHOLDS[variant] | |
| config.setdefault("data", {})["label_threshold"] = threshold | |
| config["data"]["relabel_from_future_returns"] = True | |
| result_dir = output_root / variant | |
| config["results_dir"] = str(result_dir) | |
| config["stocks"] = list(STOCKS) | |
| config["methods"] = list(METHODS) | |
| config["settings"] = ["from_scratch", "fine_tune"] | |
| config["seeds"] = list(SEEDS) | |
| config["label_budgets"] = [0.5] | |
| config["purpose"] = ( | |
| "Exploratory five-stock, three-seed prediction retraining under " | |
| f"label threshold {threshold:g}; the pooled checkpoint is fixed." | |
| ) | |
| result_dir.mkdir(parents=True, exist_ok=True) | |
| (result_dir / "resolved_config.json").write_text( | |
| json.dumps(config, indent=2) + "\n", | |
| encoding="utf-8", | |
| ) | |
| total = len(STOCKS) * len(METHODS) * len(SEEDS) | |
| completed = 0 | |
| for stock in STOCKS: | |
| for method in METHODS: | |
| for seed in SEEDS: | |
| completed += 1 | |
| print( | |
| f"THRESHOLD {variant} {completed}/{total} " | |
| f"{stock} {method} seed={seed}", | |
| flush=True, | |
| ) | |
| run_one( | |
| config, | |
| stock, | |
| method, | |
| seed, | |
| 0.5, | |
| config["settings"], | |
| force=args.force, | |
| ) | |
| summarize(config, f"label_{variant}") | |
| if __name__ == "__main__": | |
| main() | |