tokenizer_config.json

{
  "model_max_length": 1024,
  "bos_token": "<s>",
  "eos_token": "</s>",
  "unk_token": "<unk>",
  "pad_token": "<pad>",
  "additional_special_tokens": [
    "<|user|>",
    "<|assistant|>",
    "<|system|>",
    "<|tool|>"
  ],
  "chat_template": "{% set ns = namespace(system_message='') %}\n{% for message in messages %}\n{% if message['role'] == 'system' and ns.system_message == '' %}\n{% set ns.system_message = message['content'] %}\n{% endif %}\n{% endfor %}\n{% if ns.system_message == '' %}\n{% set ns.system_message = \"You are Viena, a concise Vietnamese-first assistant. Return only the final answer. Do not reveal reasoning.\" %}\n{% endif %}\n<|system|>\n{{ ns.system_message }}\n{% set ns2 = namespace(consumed=false) %}\n{% for message in messages %}\n{% if message['role'] == 'system' and not ns2.consumed %}\n{% set ns2.consumed = true %}\n{% else %}\n<|{{ message['role'] }}|>\n{{ message['content'] }}\n{% endif %}\n{% endfor %}\n{% if add_generation_prompt %}<|assistant|>\n{% endif %}",
  "tokenizer_class": "LlamaTokenizer"
}