from . import indexed_dataset from .bert_tokenization import FullTokenizer as FullBertTokenizer