tokenizer_config.json

{
  "clean_up_tokenization_spaces": true,
  "model_max_length": 512,
  "padding_side": "left",
  "tokenizer_class": "PreTrainedTokenizerFast"
}