File size: 1,754 Bytes
ba82887 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 | """Feed label-free JSONL to a local adapter; save validated JSON predictions.
Example: python run_model.py --output predictions.json -- python adapter.py
The adapter reads question objects as JSONL on stdin and emits one JSONL object
{id, prediction} per question on stdout. Logs belong on stderr. This is a local
process launcher, not a sandbox; inspect the adapter and disable network yourself.
"""
import argparse
import json
from pathlib import Path
import subprocess
from benchmark import DATASET, normalise_predictions, read_rows, dump
def main():
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--questions", default=str(DATASET / "questions/test.jsonl"))
parser.add_argument("--output", default="predictions.json")
parser.add_argument("--timeout", type=float, default=600)
parser.add_argument("command", nargs=argparse.REMAINDER)
args = parser.parse_args()
command = args.command[1:] if args.command[:1] == ["--"] else args.command
if not command:
parser.error("Provide an adapter command after --")
rows = read_rows(args.questions)
if any("answer" in r or "program" in r for r in rows):
parser.error("Use label-free questions/, not data/ or audit.jsonl")
result = subprocess.run(command, input="".join(json.dumps(r) + "\n" for r in rows),
capture_output=True, text=True, encoding="utf-8", timeout=args.timeout, check=True)
predictions = [json.loads(line) for line in result.stdout.splitlines() if line.strip()]
normalise_predictions(predictions, rows)
dump(args.output, predictions)
print(f"Validated {len(predictions)} predictions; wrote {Path(args.output).resolve()}")
if __name__ == "__main__":
main()
|