tokenizer_config.json

{
  "clean_up_tokenization_spaces": true,
  "eos_token": "</s>",
  "model_max_length": 2048,
  "pad_token": "<pad>",
  "tokenizer_class": "PreTrainedTokenizerFast"
}