config.json

{
    "model_type": "nanogpt",
    "architectures": [
        "NanoGPTChat"
    ],
    "auto_map": {
        "AutoConfig": "configuration_nanogpt.NanoGPTConfig",
        "AutoModel": "modeling_nanogpt.NanoGPTChat",
        "AutoModelForCausalLM": "modeling_nanogpt.NanoGPTChat",
        "AutoTokenizer": "tokenizer_nanogpt.NanoGPTChatTokenizer"
    },
    "bos_token": "<|bos|>",
    "eos_token": "<|assistant_end|>",
    "pad_token": "<|assistant_end|>",
    "sequence_len": 2048,
    "vocab_size": 65536,
    "n_layer": 20,
    "n_head": 10,
    "n_kv_head": 10,
    "n_embd": 1280,
    "chat_template": "{% if messages[0]['role'] == 'system' %}<|bos|><|user_start|>{{ messages[0]['content'] }}\n\n{{ messages[1]['content'] }}<|user_end|>{% set messages = messages[2:] %}{% else %}<|bos|>{% endif %}{% for message in messages %}{% if loop.index0 % 2 == 0 %}<|user_start|>{{ message['content'] }}<|user_end|>{% else %}<|assistant_start|>{{ message['content'] }}<|assistant_end|>{% endif %}{% endfor %}"
}