parse_args.py

from __future__ import unicode_literals, print_function, division
import argparse
from transformers import (
    SchedulerType,
)

def parse_args():
    parser = argparse.ArgumentParser(description="Finetune a transformers model on a text classification task")
    parser.add_argument(
        "--dataset_name",
        type=str,
        default=None,
        help="The name of the dataset to use (via the datasets library).",
    )
    parser.add_argument(
        "--dataset_config_name",
        type=str,
        default=None,
        help="The configuration name of the dataset to use (via the datasets library).",
    )
    parser.add_argument(
        "--ignore_pad_token_for_loss",
        type=bool,
        default=True,
        help="Whether to ignore the tokens corresponding to " "padded labels in the loss computation or not.",
    )
    parser.add_argument(
        "--max_source_length",
        type=int,
        default=1024,
        help="The maximum total input sequence length after "
        "tokenization.Sequences longer than this will be truncated, sequences shorter will be padded.",
    )
    parser.add_argument(
        "--source_prefix",
        type=str,
        default="",
        help="A prefix to add before every source text " "(useful for T5 models).",
    )
    parser.add_argument(
        "--overwrite_cache", type=bool, default=None, help="Overwrite the cached training and evaluation sets"
    )
    parser.add_argument(
        "--max_target_length",
        type=int,
        default=142,
        help="The maximum total sequence length for target text after "
        "tokenization. Sequences longer than this will be truncated, sequences shorter will be padded."
        "during ``evaluate`` and ``predict``.",
    )
    parser.add_argument(
        "--max_length",
        type=int,
        default=142,
        help=(
            "The maximum total input sequence length after tokenization. Sequences longer than this will be truncated,"
            " sequences shorter will be padded if `--pad_to_max_lengh` is passed."
        ),
    )
    parser.add_argument(
        "--pad_to_max_length",
        action="store_true",
        help="If passed, pad all samples to `max_length`. Otherwise, dynamic padding is used.",
    )
    parser.add_argument(
        "--base_model_pretrained_name",
        type=str,
        help="Path to pretrained model or model identifier from huggingface.co/models.",
        required=True,
    )
    parser.add_argument("--dataset_cache_dir",
        type=str,
        default=None,
        help="Dataset cache directory",)
    parser.add_argument("--pretrained_model_cache_dir",
        type=str,
        default=None,
        help="Pretrained model cache directory",)
    parser.add_argument(
        "--tokenizer_name",
        type=str,
        default=None,
        help="Pretrained tokenizer name or path if not the same as model_name",
    )
    parser.add_argument(
        "--text_column",
        type=str,
        default=None,
        help="The name of the column in the datasets containing the full texts (for summarization).",
    )
    parser.add_argument(
        "--summary_column",
        type=str,
        default=None,
        help="The name of the column in the datasets containing the summaries (for summarization).",
    )
    parser.add_argument(
        "--use_slow_tokenizer",
        action="store_true",
        help="If passed, will use a slow tokenizer (not backed by the 🤗 Tokenizers library).",
    )
    parser.add_argument(
        "--bilinear_dim",
        type=int,
        default=1024,
        help="Batch size (per device) for the training dataloader.",
    )
    parser.add_argument(
        "--per_device_train_batch_size",
        type=int,
        default=4,
        help="Batch size (per device) for the training dataloader.",
    )
    parser.add_argument(
        "--learning_rate",
        type=float,
        default=5e-5,
        help="Initial learning rate (after the potential warmup period) to use.",
    )
    parser.add_argument("--weight_decay", type=float, default=0.0, help="Weight decay to use.")
    parser.add_argument("--num_train_epochs", type=int, default=3, help="Total number of training epochs to perform.")
    parser.add_argument(
        "--max_train_steps",
        type=int,
        default=None,
        help="Total number of training steps to perform. If provided, overrides num_train_epochs.",
    )
    parser.add_argument(
        "--gradient_accumulation_steps",
        type=int,
        default=8,
        help="Number of updates steps to accumulate before performing a backward/update pass.",
    )
    parser.add_argument(
        "--lr_scheduler_type",
        type=SchedulerType,
        default="linear",
        help="The scheduler type to use.",
        choices=["linear", "cosine", "cosine_with_restarts", "polynomial", "constant", "constant_with_warmup"],
    )
    parser.add_argument(
        "--num_warmup_steps", type=int, default=0, help="Number of steps for the warmup in the lr scheduler."
    )
    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
    parser.add_argument("--n_train_data_samples", type=int, default=None, help="Specify training dataset samples.")
    parser.add_argument('--log_freq', type=int, required=False, default=20,
                        help='Log frequency on screen.')

    args = parser.parse_args()

    # Sanity checks
    if args.dataset_name is None:
        raise ValueError("Need either a dataset name or a training/validation file.")

    return args