{
    "model_id": "glm4v_moe",
    "init_from_hf_path": "./ckpt/hf_path/GLM-4.5V",
    "image_encoder": {
        "vision_encoder": {
            "model_id": "glm4v_moe",
            "num_layers": 24,
            "hidden_size": 1536,
            "num_attention_heads": 12,
            "freeze": true,
            "attn_implementation": "sdpa"
        },
        "vision_projector": {
            "model_id": "lnmlp",
            "num_layers": 1,
            "freeze": true
        }
    },
    "text_decoder": {
        "model_id": "glm_4v_moe_text",
        "num_layers": 46,
        "hidden_size": 12288,
        "num_attention_heads": 96,
        "max_position_embeddings": 65536,
        "freeze": false,
        "use_npu_fused_moe": true,
        "attn_implementation": "sdpa",
        "activation_offload": false
    },
    "loss_cfg": {
        "compute_mode": "default",
        "chunk_size": 1024,
        "loss_type": "per_token_loss"
    },
    "patch": {
        "clip_grad_async": true,
        "scale_grad": true
    }
}