tokenizer_config.json

{
  "add_prefix_space": false,
  "backend": "tokenizers",
  "bos_token": "<|im_start|>",
  "clean_up_tokenization_spaces": false,
  "eos_token": "<|im_end|>",
  "errors": "replace",
  "extra_special_tokens": [
    "<|im_start|>",
    "<|im_end|>"
  ],
  "is_local": true,
  "model_max_length": 2048,
  "pad_token": "<|im_end|>",
  "tokenizer_class": "GPT2Tokenizer",
  "unk_token": "<|endoftext|>",
  "vocab_size": 49152
}