{
    "model_id": "qwen3_vl",
    "img_context_token_id": 151655,
    "video_token_id": 151656,
    "vision_start_token_id": 151652,
    "text_decoder": {
        "model_id": "qwen3_lm",
        "add_qkv_bias": false,
        "attention_dropout": 0.0,
        "params_dtype": "bf16",
        "bf16": true,
        "kv_channels": 128,
        "activation_func": "silu",
        "hidden_size": 4096,
        "ffn_hidden_size": 12288,
        "max_position_embeddings": 262144,
        "num_attention_heads": 32,
        "num_layers": 36,
        "num_query_groups": 8,
        "layernorm_epsilon": 1e-06,
        "rope_scaling": {
            "mrope_interleaved": true,
            "mrope_section": [
                24,
                20,
                20
            ],
            "rope_type": "default"
        },
        "rope_theta": 5000000,
        "vocab_size": 151936,
        "pipeline_num_layers": [6, 10, 10, 10],
        "seq_length": 1024,
        "untie_embeddings_and_output_weights": true,
        "disable_bias_linear": true,
        "init_method_std": 0.01,
        "hidden_dropout": 0.0,
        "position_embedding_type": "mrope",
        "normalization": "RMSNorm",
        "use_fused_rotary_pos_emb": true,
        "attention_softmax_in_fp32": true,
        "parallel_output": true,
        "group_query_attention": true,
        "moe_grouped_gemm": false,
        "gated_linear_unit": true,
        "add_bias_linear":false,
        "use_remove_padding": false,
        "qk_layernorm": true,
        "use_qk_norm": true,
        "sequence_parallel": false,
        "recompute_granularity": "full",
        "recompute_method": "block",
        "recompute_num_layers": 4,
        "use_te": false
    },
    "image_encoder": {
        "vision_encoder": {
            "model_id": "qwen3vit",
            "deepstack_visual_indexes": [
                8,
                16,
                24
            ],
            "num_layers": 27,
            "activation_func": "gelu_pytorch_tanh",
            "hidden_size": 1152,
            "in_channels": 3,
            "ffn_hidden_size": 4304,
            "num_attention_heads": 16,
            "max_position_embeddings": 2304,
            "llm_hidden_size": 4096,
            "patch_size": 16,
            "spatial_merge_size": 2,
            "temporal_patch_size": 2,
            "add_bias_linear": true,
            "add_bias_conv": true,
            "gated_linear_unit": false,
            "bias_activation_fusion": false,
            "hidden_dropout": 0.0,
            "attention_dropout": 0.0,
            "layernorm_epsilon": 1e-06,
            "normalization": "LayerNorm",
            "fp16": false,
            "bf16": true,
            "params_dtype": "bf16",
            "freeze": true,
            "use_fused_rotary_pos_emb": true,
            "post_layer_norm": false,
            "pipeline_num_layers": [27, 0, 0, 0],
            "intermediate_size": 4304,
            "tokens_per_second": 2,
            "attention_mask_type": "general"
        },
        "vision_projector": {
            "model_id": "lnmlp",
            "num_layers": 1,
            "gated_linear_unit": false,
            "bias_activation_fusion": false,
            "add_bias_linear": true,
            "input_size": 1152,
            "hidden_size": 4096,
            "ffn_hidden_size": 4608,
            "activation_func": "gelu",
            "bf16": true,
            "params_dtype": "bf16",
            "freeze": true,
            "layernorm_epsilon": 1e-06,
            "normalization": "LayerNorm",
            "use_postshuffle_norm": false
        }
    },

    "text_encoder": null,
    "video_encoder": null
}