{ "tokenizer_class": "RnaTokenizer", "clean_up_tokenization_spaces": true }