tokenizer.json

from transformers import BertTokenizer

tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
tokenizer.save_pretrained('./hf_model')