{
  "tokenizer_class": "PreTrainedTokenizerFast",
  "model_max_length": 512,
  "eos_token": "<|endoftext|>"
}