def create_parser()

in variant-prediction/predict.py [0:0]

• 57 lines of code
• 1 McCabe index (conditional complexity)

def create_parser():
    parser = argparse.ArgumentParser(
        description="Label a deep mutational scan with predictions from an ensemble of ESM-1v models."  # noqa
    )

    # fmt: off
    parser.add_argument(
        "--model-location",
        type=str,
        help="PyTorch model file OR name of pretrained model to download (see README for models)",
        nargs="+",
    )
    parser.add_argument(
        "--sequence",
        type=str,
        help="Base sequence to which mutations were applied",
    )
    parser.add_argument(
        "--dms-input",
        type=pathlib.Path,
        help="CSV file containing the deep mutational scan",
    )
    parser.add_argument(
        "--mutation-col",
        type=str,
        default="mutant",
        help="column in the deep mutational scan labeling the mutation as 'AiB'"
    )
    parser.add_argument(
        "--dms-output",
        type=pathlib.Path,
        help="Output file containing the deep mutational scan along with predictions",
    )
    parser.add_argument(
        "--offset-idx",
        type=int,
        default=0,
        help="Offset of the mutation positions in `--mutation-col`"
    )
    parser.add_argument(
        "--scoring-strategy",
        type=str,
        default="wt-marginals",
        choices=["wt-marginals", "pseudo-ppl", "masked-marginals"],
        help=""
    )
    parser.add_argument(
        "--msa-path",
        type=pathlib.Path,
        help="path to MSA (required for MSA Transformer)"
    )
    parser.add_argument(
        "--msa-samples",
        type=int,
        default=400,
        help="number of sequences to randomly sample from the MSA"
    )
    # fmt: on
    parser.add_argument("--nogpu", action="store_true", help="Do not use GPU even if available")
    return parser