I am trying to fine tune a Roberta model after adding some special tokens to its tokenizer:
special_tokens_dict = {\'additional_special_tokens\': [\'[Tok1