mlc-chat-config.json

{
    "model_type": "llama",
    "quantization": "q4f16_0",
    "model_config": {
        "hidden_size": 4096,
        "intermediate_size": 14336,
        "num_attention_heads": 32,
        "num_hidden_layers": 32,
        "rms_norm_eps": 1e-5,
        "vocab_size": 102400,
        "position_embedding_base": 500000.0,
        "context_window_size": 2048,
        "prefill_chunk_size": 2048,
        "num_key_value_heads": 8,
        "head_dim": 128,
        "tensor_parallel_shards": 1,
        "max_batch_size": 80
    },
    "vocab_size": 102400,
    "context_window_size": 2048,
    "sliding_window_size": -1,
    "prefill_chunk_size": 2048,
    "attention_sink_size": -1,
    "tensor_parallel_shards": 1,
    "mean_gen_len": 128,
    "max_gen_len": 512,
    "shift_fill_factor": 0.3,
    "temperature": 0,
    "presence_penalty": 0.0,
    "frequency_penalty": 0.0,
    "repetition_penalty": 1.0,
    "top_p": 0.95,
    "conv_template": {
        "name": "exaone_llama",
        "system_template": "{system_message}",
        "system_message": "",
        "system_prefix_token_ids": [],
        "add_role_after_system_message": true,
        "roles": {
            "user": "[BOS][INST] ",
            "assistant": "[/INST]"
        },
        "role_templates": {
            "user": "{user_message}",
            "assistant": "{assistant_message}",
            "tool": "{tool_message}"
        },
        "messages": [],
        "seps": [" "],
        "role_content_sep": "",
        "role_empty_sep": "",
        "stop_str": ["[|endofturn|]"],
        "stop_token_ids": [361],
        "function_string": "",
        "use_function_calling": false,
        "image_token_index": -1
    },
    "pad_token_id": 0,
    "bos_token_id": 1,
    "eos_token_id": 361,
    "tokenizer_files": [
        "tokenizer.json",
        "vocab.json",
        "merges.txt",
        "tokenizer_config.json"
    ],
    "version": "0.1.0"
}