tokenizer_config.json

{
  "tokenizer_class": "PreTrainedTokenizerFast",
  "model_max_length": 512
}