llama-3-8b-chatml/special_tokens_map.json

{
  "bos_token": "<|begin_of_text|>",
  "eos_token": "<|end_of_text|>"
}