tokenizer_config.json

{
  "add_prefix_space": false,
  "model_max_length": 512,
  "tokenizer_class": "RobertaTokenizer"
}