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