vocab/chinese_llama/tokenizer/special_tokens_map.json

{
  "bos_token": "<s>",
  "eos_token": "</s>",
  "unk_token": "<unk>"
}