INNER CODE UNIT · Python

main

Future-House/BixBench · generate_zeroshot_evals.py:133

async def main():
    logger.info("Starting evaluation...")
    args = parse_args()
    output_file = (
        args.output_file
        if args.output_file is not None
        else f"results_{args.answer_mode}_{args.with_refusal}_{args.model}_{args.temperature}.csv"
    )
    if args.local_csv is None:
        _hf_login()
        dataset = load_dataset(HF_URL)[args.dataset_split].to_pandas()  # type: ignore[index]
    else:
        dataset = pd.read_csv(args.local_csv)
    dataset = cast("pd.DataFrame", dataset)
    if args.num_examples > 0:
        dataset = dataset.head(args.num_examples)

    zeroshot_agent = ZeroshotBaseline(

View source record →