config.json

{
  "model_type": "bailing_moe_v3_vl",
  "architectures": [
    "BailingMoeV3VLForConditionalGeneration"
  ],
  "auto_map": {
    "AutoConfig": "configuration_bailing_moe_v3_vl.BailingMoeV3VLConfig",
    "AutoModelForCausalLM": "modeling_bailing_moe_v3_vl.BailingMoeV3VLForConditionalGeneration"
  },
  "torch_dtype": "bfloat16",
  "transformers_version": "4.56.2",
  "tie_word_embeddings": false,
  "norm_query_embeds": false,
  "image_token_id": 157157,
  "video_token_id": 156909,
  "vision_start_token_id": 157158,
  "vision_end_token_id": 157159,
  "mrope_section": [
    8,
    12,
    12
  ],
  "text_config": {
    "image_patch_token": 157157,
    "video_patch_token": 156909,
    "image_start_token": 157158,
    "video_start_token": 157160,
    "num_hidden_layers": 42,
    "hidden_size": 2560,
    "intermediate_size": 6144,
    "eos_token_id": 156895,
    "pad_token_id": 156892,
    "first_k_dense_replace": 2,
    "max_position_embeddings": 131072,
    "moe_intermediate_size": 768,
    "num_experts_per_tok": 8,
    "num_attention_heads": 32,
    "q_lora_rank": null,
    "kv_lora_rank": 512,
    "qk_nope_head_dim": 128,
    "qk_rope_head_dim": 64,
    "v_head_dim": 128,
    "num_experts": 512,
    "num_key_value_heads": 32,
    "rope_theta": 6000000,
    "rms_norm_eps": 0.000001,
    "head_dim": 128,
    "vocab_size": 157184,
    "partial_rotary_factor": 0.5,
    "router_dtype": "fp32",
    "moe_router_enable_expert_bias": true,
    "routed_scaling_factor": 2.5,
    "n_group": 8,
    "topk_group": 4,
    "use_qk_norm": true,
    "score_function": "sigmoid",
    "moe_shared_expert_intermediate_size": 768,
    "layer_group_size": 6,
    "num_kv_heads_for_linear_attn": 0,
    "group_norm_size": 1,
    "linear_silu": true,
    "rotary_dim": 64,
    "use_mla_nope": false,
    "short_conv_kernel_size": 4,
    "use_nGPT": false,
    "scale_router_input": false,
    "value_norm": false,
    "up_proj_norm": false,
    "gated_attention_proj_granularity_type": "head_wise",
    "mtp_use_kda": false,
    "no_kda_lora": true,
    "use_kda_lora": false,
    "kda_safe_gate": true,
    "kda_lower_bound": -5.0,
    "norm_topk_prob": true,
    "expert_swiglu_limit_list": [
      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
      0, 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4
    ],
    "share_expert_swiglu_limit_list": [
      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
      0, 0, 0, 0, 0, 0, 5, 5, 5, 5, 5, 5, 7, 7
    ]
  },
  "vision_config": {
    "model_type": "qwen3_moe_vit",
    "depth": 27,
    "hidden_size": 1152,
    "hidden_act": "gelu_pytorch_tanh",
    "intermediate_size": 4304,
    "num_heads": 16,
    "in_channels": 3,
    "patch_size": 16,
    "spatial_merge_size": 2,
    "temporal_patch_size": 2,
    "out_hidden_size": 4096,
    "num_position_embeddings": 2304,
    "initializer_range": 0.02,
    "disable_merger_proj": true
  },
  "quantization_config": {
    "activation_scheme": "dynamic",
    "fmt": "e4m3",
    "quant_method": "fp8",
    "weight_block_size": [
      128,
      128
    ],
    "modules_to_not_convert": [
      "model.visual",
      "linear_proj",
      "lm_head",
      "model.word_embeddings",
      "b_proj",
      "mlp.gate",
      "model.layers.5.attention.g_proj",
      "model.layers.11.attention.g_proj",
      "model.layers.17.attention.g_proj",
      "model.layers.23.attention.g_proj",
      "model.layers.29.attention.g_proj",
      "model.layers.35.attention.g_proj",
      "model.layers.41.attention.g_proj"
    ]
  }
}