vocab/bert_base_cased/__init__.py

from transformers import BertTokenizer
tokenizer = BertTokenizer.from_pretrained('bert-base-cased')