| |
| """Convert Plan D RAG predict output (2618 rows aligned to Test_aug_sbert.csv) to Zindi CSV. |
| Plan D predicts 1:1 with Test.csv rows (no canonical dedup), so just zip preds with Test.csv IDs. |
| """ |
| import json, argparse, pathlib |
| import pandas as pd |
|
|
| WORK = pathlib.Path('/mnt/msrh/Magic_submission') |
| SUFFIX = ' This answer provides accurate and complete health information.' |
|
|
|
|
| def main(): |
| ap = argparse.ArgumentParser() |
| ap.add_argument('--pred_jsonl', required=True) |
| ap.add_argument('--test_csv', default=str(WORK / 'data' / 'Test.csv')) |
| ap.add_argument('--out_csv', required=True) |
| ap.add_argument('--no_suffix', action='store_true') |
| args = ap.parse_args() |
|
|
| preds = [] |
| with open(args.pred_jsonl) as f: |
| for line in f: |
| r = json.loads(line) |
| p = (r.get('predict') or r.get('prediction') or '').strip() |
| preds.append(p) |
|
|
| test = pd.read_csv(args.test_csv) |
| assert len(preds) == len(test), f"len mismatch: {len(preds)} preds vs {len(test)} test" |
|
|
| rows = [] |
| for tid, pred in zip(test['ID'], preds): |
| ans = pred + ('' if args.no_suffix else SUFFIX) |
| rows.append({'ID': tid, 'TargetRLF1': ans, 'TargetR1F1': ans, 'TargetLLM': ans}) |
| pd.DataFrame(rows, columns=['ID', 'TargetRLF1', 'TargetR1F1', 'TargetLLM']).to_csv(args.out_csv, index=False) |
| print(f"wrote {args.out_csv} ({len(rows)} rows)") |
|
|
|
|
| if __name__ == '__main__': |
| main() |
|
|
