tokenizer_config.json

{
  "bos_token": "<|startoftext|>",
  "clean_up_tokenization_spaces": true,
  "eos_token": "<|eos|>",
  "model_max_length": 262144,
  "pad_token": "<|pad|>",
  "tokenizer_class": "PreTrainedTokenizerFast"
}