added_tokens.json

{
  "<unk>": 4900,
  "<|endoftext|>": 4901,
  "<|im_start|>": 4902,
  "<|im_end|>": 4903,
  "<|object_ref_start|>": 4904,
  "<|object_ref_end|>": 4905,
  "<|box_start|>": 4906,
  "<|box_end|>": 4907,
  "<|quad_start|>": 4908,
  "<|quad_end|>": 4909,
  "<|vision_start|>": 4910,
  "<|vision_end|>": 4911,
  "<|vision_pad|>": 4912,
  "<|image_pad|>": 4913,
  "<|video_pad|>": 4914,
  "<image>": 4915,
  "</image>": 4916,
  "<ref>": 4917,
  "</ref>": 4918,
  "<box>": 4919,
  "</box>": 4920,
  "<quad>": 4921,
  "</quad>": 4922,
  "<point>": 4923,
  "</point>": 4924,
  "<slice>": 4925,
  "</slice>": 4926,
  "<image_id>": 4927,
  "</image_id>": 4928,
  "<unit>": 4929,
  "</unit>": 4930,
  "<asr>": 4931,
  "</asr>": 4932,
  "<query>": 4933,
  "</query>": 4934,
  "<|audio_start|>": 4935,
  "<|audio|>": 4936,
  "<|audio_end|>": 4937,
  "<|spk_bos|>": 4938,
  "<|spk|>": 4939,
  "<|spk_eos|>": 4940,
  "<|tts_bos|>": 4941,
  "<|tts_eos|>": 4942,
  "<|listen|>": 4943,
  "<|speak|>": 4944,
  "<|interrupt|>": 4945,
  "<|vad_start|>": 4946,
  "<|vad_end|>": 4947,
  "<reserved_43>": 4948,
  "<reserved_53>": 4949
}