"""packing huggingface datasets"""
import sys
import argparse
import os
import time
from mindformers.tools import MindFormerConfig, logger
from mindformers.dataset.dataloader.hf_dataloader import HFDataLoader, process_legacy_args
from mindformers.core.context.build_context import MFContextOperator
class ProcessHFDataLoader(HFDataLoader):
"""HF dataset process module"""
def __new__(cls, **kwargs):
kwargs = process_legacy_args(**kwargs)
config = cls._build_config(**kwargs)
dataset = cls.load_dataset(config)
cls.process_dataset(config, dataset)
def prepare_args():
"""prepare arguments for packing dataset"""
parser = argparse.ArgumentParser()
parser.add_argument('--config',
type=str,
required=True,
help='A path of config file containing CommonDataLoader.')
parser.add_argument('--save_path',
default="./packed_data",
type=str,
help='A directory to save packed dataset.')
parser.add_argument('--register_path',
type=str,
default=None,
help='The register path of outer API. '
'This is usually the parent path of the Python file where the outer API is located.'
'This configuration can be ignored if it does not involve registering the outer API.')
args = parser.parse_args()
if args.register_path is not None:
work_path = os.path.dirname(os.path.abspath(__file__))
work_path = '/'.join(work_path.split('/')[:-3])
if not os.path.isabs(args.register_path):
args.register_path = os.path.join(work_path, args.register_path)
os.environ["REGISTER_PATH"] = args.register_path
if args.register_path not in sys.path:
sys.path.append(args.register_path)
return args
def main():
"""main processing function"""
start_time = time.time()
args = prepare_args()
config = MindFormerConfig(args.config)
MFContextOperator(config)
dataloader_config = config.train_dataset.data_loader
dataloader_config.handler.append(dict(type='save_to_disk', dataset_path=args.save_path))
ProcessHFDataLoader(**dataloader_config)
end_time = time.time()
logger.info(f"Processed datasets saved in {args.save_path}, spend {end_time - start_time:.4f}s.")
if __name__ == '__main__':
main()