def add_model_args()

in elq/common/params.py [0:0]

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

    def add_model_args(self, args=None):
        """
        Add model args.
        """
        parser = self.add_argument_group("Model Arguments")
        parser.add_argument(
            "--max_seq_length",
            default=256,
            type=int,
            help="The maximum total input sequence length after WordPiece tokenization. \n"
            "Sequences longer than this will be truncated, and sequences shorter \n"
            "than this will be padded.",
        )
        parser.add_argument(
            "--max_context_length",
            default=128,
            type=int,
            help="The maximum total context input sequence length after WordPiece tokenization. \n"
            "Sequences longer than this will be truncated, and sequences shorter \n"
            "than this will be padded.",
        )
        parser.add_argument(
            "--max_cand_length",
            default=128,
            type=int,
            help="The maximum total label input sequence length after WordPiece tokenization. \n"
            "Sequences longer than this will be truncated, and sequences shorter \n"
            "than this will be padded.",
        ) 
        parser.add_argument(
            "--path_to_model",
            default=None,
            type=str,
            required=False,
            help="The full path to the model to load.",
        )
        parser.add_argument(
            "--bert_model",
            default="bert-base-uncased",
            type=str,
            help="Bert pre-trained model selected in the list: bert-base-uncased, "
            "bert-large-uncased, bert-base-cased, bert-base-multilingual, bert-base-chinese.",
        )
        parser.add_argument(
            "--pull_from_layer", type=int, default=-1, help="Layers to pull from BERT",
        )
        parser.add_argument(
            "--lowercase",
            action="store_false",
            help="Whether to lower case the input text. True for uncased models, False for cased models.",
        )
        parser.add_argument("--context_key", default="context", type=str)
        parser.add_argument("--title_key", default="entity", type=str)
        parser.add_argument(
            "--out_dim", type=int, default=1, help="Output dimention of bi-encoders.",
        )
        parser.add_argument(
            "--add_linear",
            action="store_true",
            help="Whether to add an additonal linear projection on top of BERT.",
        )
        parser.add_argument(
            "--data_path",
            default="data/zeshel",
            type=str,
            help="The path to the train data.",
        )
        parser.add_argument(
            "--output_path",
            default=None,
            type=str,
            required=True,
            help="The output directory where generated output file (model, etc.) is to be dumped.",
        )
        parser.add_argument(
            "--mention_aggregation_type",
            default=None,
            type=str,
            help="Type of mention aggregation (None to just use [CLS] token, "
            "'all_avg' to average across tokens in mention, 'fl_avg' to average across first/last tokens in mention, "
            "'{all/fl}_linear' for linear layer over mention, '{all/fl}_mlp' to MLP over mention)",
        )
        parser.add_argument(
            "--no_mention_bounds",
            dest="no_mention_bounds",
            action="store_true",
            default=False,
            help="Don't add tokens around target mention. MUST BE FALSE IF 'mention_aggregation_type' is NONE",
        )
        parser.add_argument(
            "--mention_scoring_method",
            dest="mention_scoring_method",
            default="qa_linear",
            type=str,
            help="Method for generating/scoring mentions boundaries (options: 'qa_mlp', 'qa_linear', 'BIO')",
        )
        parser.add_argument(
            "--max_mention_length",
            dest="max_mention_length",
            default=10,
            type=int,
            help="Maximum length of span to consider as candidate mention",
        )