push_to_hub.py

from pathlib import Path
from datasets import load_dataset, load_from_disk
from dataclasses import dataclass, field
from huggingface_hub import HfApi
from transformers import AutoModel, AutoTokenizer, HfArgumentParser
from typing import Optional, List


@dataclass
class DownloadArgs:
    model_cache_dir: str = field(
        default='/share/LMs',
        metadata={'help': 'Default path to save language models'}
    )
    model_name_or_path: Optional[str] = field(
        default=None,
        metadata={'help': 'Path to pretrained model or model identifier from huggingface.co/models'}
    )
    dataset_cache_dir: str = field(
        default='/share/peitian/Data/Datasets/huggingface',
        metadata={'help': 'Default path to save huggingface datasets'}
    )
    dataset_name_or_path: Optional[str] = field(
        default=None,
        metadata={'help': 'Dataset name'}
    )
    data_files: Optional[dict] = field(
        default=None,
        metadata={'help': 'Data files for json dataset.'}
    )
    dataset_from_disk: bool = field(
        default=False,
        metadata={'help': 'Load dataset from disk?'}
    )
    
    file: Optional[str] = field(
        default=None,
        metadata={'help': 'File to upload.'}
    )
    file_in_repo: Optional[str] = field(
        default=None,
        metadata={'help': 'File name in repository.'}
    )
    
    hub_name: Optional[str] = field(
        default=None,
        metadata={'help': 'Name of the huggingface repo.'}
    )
    
    revision: str = field(
        default=None,
        metadata={'help': 'Remote code revision'}
    )
    resume_download: bool = field(
        default=True,
        metadata={'help': 'Resume downloading'}
    )
    def __post_init__(self):
        # folder or model not exists
        if self.model_name_or_path is not None:
            tokenizer = AutoTokenizer.from_pretrained(self.model_name_or_path, cache_dir=self.model_cache_dir, trust_remote_code=True)
            model = AutoModel.from_pretrained(self.model_name_or_path, cache_dir=self.model_cache_dir, trust_remote_code=True)
            # use loop to force success upload
            while 1:
                try:
                    tokenizer.push_to_hub(self.hub_name)
                    break
                except:
                    pass
            while 1:
                try:
                    model.push_to_hub(self.hub_name)
                    break
                except:
                    pass
            
        if self.dataset_name_or_path is not None:
            if self.dataset_from_disk:
                dataset = load_from_disk(self.dataset_name_or_path)
            else:
                dataset = load_dataset(self.dataset_name_or_path, data_files=self.data_files, cache_dir=self.dataset_cache_dir)
            # use loop to force success upload
            while 1:
                try:
                    dataset.push_to_hub(self.hub_name)
                    break
                except:
                    pass
        
        if self.file is not None:
            api = HfApi()
            if self.file_in_repo is None:
                self.file_in_repo = Path(self.file).name
            api.upload_file(
                path_or_fileobj=self.file,
                path_in_repo=self.file_in_repo,
                repo_id=self.hub_name,
                repo_type="dataset",
            )


if __name__ == "__main__":
    parser = HfArgumentParser([DownloadArgs])
    args, = parser.parse_args_into_dataclasses()