special_tokens_dict = {'additional_special_tokens': ['[C1]','[C2]','[C3]','[C4]']} num_added_toks = tokenizer.add_special_tokens(special_tokens_dict) model.resize_token_embeddings(len(tokenizer))