{
"img_context_token_id": 151655,
"vision_start_token_id": 151652,
"image_encoder": {
"vision_encoder": {
"model_id": "qwen2vit",
"num_layers": 32,
"pipeline_num_layers": [32],
"expert_model_parallel_size": 1,
"num_moe_experts": null,
"hidden_size": 1280,
"ffn_hidden_size": 3420,
"intermediate_size": 3584,
"num_attention_heads": 16,
"gated_linear_unit": true,
"hidden_dropout": 0.0,
"attention_dropout": 0.0,
"in_channels": 3,
"patch_size": 14,
"spatial_merge_size": 2,
"temporal_patch_size": 2,
"layernorm_epsilon": 1e-06,
"tokens_per_second": 2,
"window_attn_size": 112,
"fullatt_block_indexes": [
7,
15,
23,
31
],
"fp16": false,
"bf16": true,
"params_dtype": "bf16",
"activation_func": "silu",
"normalization": "RMSNorm",
"use_fused_rotary_pos_emb": true,
"post_layer_norm": false,
"freeze": true
},
"vision_projector": {
"model_id": "lnmlp",
"num_layers": 1,
"gated_linear_unit": false,
"bias_activation_fusion": false,
"expert_model_parallel_size": 1,
"num_moe_experts": null,
"add_bias_linear": true,
"input_size": 1280,
"hidden_size": 2048,
"ffn_hidden_size": 5120,
"activation_func": "gelu",
"bf16": true,
"params_dtype": "bf16",
"layernorm_epsilon": 1e-06,
"normalization": "RMSNorm",
"freeze": true
}
},
"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,
"recompute_granularity": "full",
"recompute_method": "uniform",
"recompute_num_layers": 1
},
"text_encoder": null,
"video_encoder": null
}