tokenizer_config.json

{
  "clean_up_tokenization_spaces": true,
  "model_max_length": 2048,
  "special_tokens": [
    "<|pad|>",
    "<|endoftext|>",
    "<|beginoftext|>",
    "<|unk|>"
  ],
  "tokenizer_class": "PreTrainedTokenizerFast"
}