{
    "llm_model": "llm.mnn",
    "llm_weight": "llm.mnn.weight",
    "visual_model": "visual.mnn",
    "backend_type": "cpu",
    "thread_num": 40,
    "precision": "low",
    "memory": "low",
    "sampler_type": "penalty",
    "penalty": 1.1,
    "hidden_states": true,
    "all_logits": true,
    "mllm": {
        "backend_type": "cpu",
        "thread_num": 40,
        "precision": "low",
        "memory": "low"
    }
}