{
    "img_context_token_id": 151655,
    "vision_start_token_id": 151652,
    "image_encoder": {
        "vision_encoder": {
            "model_id": "InternViT",
            "num_layers": 24,
            "hidden_size": 1024,
            "ffn_hidden_size": 4096,
            "num_attention_heads": 16,
            "num_channels": 3,
            "patch_size": 14,
            "image_size": 448,
            "add_qkv_bias": true,
            "qk_layernorm": false,
            "activation_func": "gelu",
            "normalization": "LayerNorm",
            "layernorm_epsilon": 1e-6,
            "hidden_dropout": 0.0,
            "drop_path_rate": 0.0,
            "attention_dropout": 0.0,
            "init_method_std": 0.02,
            "initializer_factor": 1.0,
            "output_hidden_states": false,
            "use_return_dict": false,
            "recompute_granularity": "full",
            "recompute_method": "uniform",
            "recompute_num_layers": 1,
            "params_dtype": "bf16",
            "post_layer_norm": false,
            "downsample_ratio": 0.5,
            "fp16": false,
            "bf16": true,
            "attention_softmax_in_fp32": false,
            "select_layer": -1,
            "ps_version": "v2",
            "freeze": true
        },
        "vision_projector": {
            "model_id": "InternVLMLP",
            "downsample_ratio": 0.5,
            "vit_hidden_size": 1024,
            "llm_hidden_size": 2048
        }
    },
    "text_decoder": {
        "model_id": "qwen3_lm",
        "num_layers": 48,
        "pipeline_num_layers": [48],
        "hidden_size": 2048,
        "ffn_hidden_size": 768,
        "moe_intermediate_size": 768,
        "num_attention_heads": 32,
        "kv_channels": 128,
        "group_query_attention": true,
        "num_query_groups": 4,
        "position_embedding_type": "mrope",
        "mrope_section": [16, 24, 24],
        "use_fused_rotary_pos_emb": true,
        "rope_scaling": null,
        "normalization": "RMSNorm",
        "layernorm_epsilon": 1e-06,
        "qk_layernorm": true,
        "activation_func": "silu",
        "gated_linear_unit": true,
        "add_bias_linear":false,
        "add_qkv_bias": false,
        "disable_bias_linear": true,
        "attention_softmax_in_fp32": true,
        "untie_embeddings_and_output_weights": true,
        "attention_dropout": 0.0,
        "init_method_std": 0.01,
        "hidden_dropout": 0.0,
        "seq_length": 1024,
        "max_position_embeddings": 40960,
        "vocab_size": 151936,
        "rope_theta": 1000000.0,
        "params_dtype": "bf16",
        "bf16": true,
        "parallel_output": true,
        "num_moe_experts": 128,
        "norm_topk_prob": true,
        "moe_router_topk": 8,
        "moe_router_load_balancing_type": "aux_loss",
        "moe_aux_loss_coeff": 0.001,
        "moe_router_score_function": "softmax",
        "moe_router_pre_softmax": false,
        "routed_scaling_factor": 2.0,
        "router_gating_in_fp32": false,
        "moe_grouped_gemm": true,
        "recompute_granularity": "full",
        "recompute_method": "uniform",
        "recompute_num_layers": 1
    },
    "text_encoder": null,
    "video_encoder": null
}