# load the hand-crafted seed pairs at data/seed/synthetic_pairs.jsonl and # split them stratified-by-category into train/val/test under data/pairs/. import argparse from pathlib import Path from cleanup.config import load_data_config from cleanup.data.download import build_and_save def main() -> None: parser = argparse.ArgumentParser() parser.add_argument("--config", default="configs/data.yaml") parser.add_argument("--out", default="data/pairs") parser.add_argument("--smoke", action="store_true", help="tiny subset for wiring tests") args = parser.parse_args() cfg = load_data_config(args.config) build_and_save(cfg, Path(args.out), smoke=args.smoke) if __name__ == "__main__": main()