tokenizer_config.json

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