HF Argparser

Standalone Argument Parser from HuggingFace Transformers.

This is a standalone version of the argument parser used in the HuggingFace Transformers library. It is a simple wrapper around the argparse library that allows you to easily define your command-line arguments for your scripts using dataclasses.

Installation

pip install hf-argparser

Usage

from dataclasses import dataclass
from hf_argparser import HfArgumentParser, HFArg

@dataclass
class AddArgs():
    x: int
    y: int

@dataclass
class OutputArgs():
    output_file: str = field(
        default=None,
        metadata={'help': 'output filename'})

    parser = HfArgumentParser([AddArgs, OutputArgs])
    
    add_args, output_args, unknown_args = parser.parse_args_into_dataclasses(
            return_remaining_strings=True)
    return add_args, output_args, unknown_args

Above code is taken from this post by @kenilc.

See the Transformers documentation for more information on how to use the HfArgumentParser.

Todo

• [] Automate pulling new changes to hf_argparser from transformers repo.
• [] Automate creating new releases on PyPi.
• [] Add tests.
• [] Improve README.

Acknowledgements

• HuggingFace Transformers :hugs: