Datasets:
File size: 821 Bytes
96a739e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 | #!/usr/bin/env python3
"""Export original model messages and case IDs, excluding labels and metadata."""
import argparse
import json
from pathlib import Path
def main():
ap = argparse.ArgumentParser(description=__doc__)
ap.add_argument('--data-file',type=Path,default=Path(__file__).parent/'sandglass_bench.jsonl')
ap.add_argument('--output',type=Path,required=True)
args = ap.parse_args()
args.output.parent.mkdir(parents=True,exist_ok=True)
with args.data_file.open(encoding='utf-8') as source,args.output.open('w',encoding='utf-8') as output:
for line in source:
if line.strip():
row=json.loads(line)
output.write(json.dumps(dict(case_id=row['case_id'],messages=row['prompt']),ensure_ascii=False)+'\n')
if __name__ == '__main__':
main()
|