tokenizer/64K/tokenizer_config.json

{
  "tokenizer_class": "PreTrainedTokenizerFast",
  "model_max_length": 1048576,
  "clean_up_tokenization_spaces": false,
  "add_bos_token": false,
  "add_eos_token": false,
  "bos_token": "<|bos|>",
  "eos_token": "<|endoftext|>",
  "pad_token": "<|pad|>",
  "unk_token": null,
  "additional_special_tokens": [
    "<|eos|>",
    "<|sep|>",
    "<|mask|>",
    "<|system|>",
    "<|user|>",
    "<|assistant|>",
    "<|tool|>",
    "<|end|>",
    "<|tool_call|>",
    "<|tool_result|>",
    "<|fim_prefix|>",
    "<|fim_suffix|>",
    "<|fim_middle|>",
    "<|image|>",
    "<|audio|>",
    "<|video|>",
    "<|code|>",
    "<|curl|>",
    "<|email|>",
    "<|cite|>",
    "<|think|>",
    "<|im_start|>",
    "<|im_end|>",
    "<|endofprompt|>",
    "<|repo_name|>",
    "<|file_sep|>",
    "<|image_start|>",
    "<|image_end|>",
    "<|image_pad|>"
  ],
  "chat_template": "{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{'<|im_start|>assistant\n'}}{% endif %}"
}