config.json

{
  "architectures": [
    "LLaDA2MoE_VLForConditionalGeneration"
  ],
  "auto_map": {
    "AutoConfig": "configuration_llada2_vl.LLaDA2VLMoEConfig",
    "AutoModel": "modeling_llada2_vl_moe.LLaDA2MoE_VLForConditionalGeneration",
    "AutoModelForCausalLM": "modeling_llada2_vl_moe.LLaDA2MoE_VLForConditionalGeneration"
  },
  "image_token_id": 157187,
  "is_composition": true,
  "model_type": "llada2vl",
  "text_config": {
    "_name_or_path": "llada2_vl_qwen2.5",
    "architectures": [
      "LLaDA2MoE_VLForConditionalGeneration"
    ],
    "attention_bias": false,
    "attention_dropout": 0.0,
    "clip_qkv": null,
    "dense_intermediate_size": -1,
    "embedding_dropout": 0.0,
    "expert_intermediate_size": -1,
    "first_k_dense_replace": 1,
    "head_dim": 128,
    "hidden_act": "silu",
    "hidden_size": 2048,
    "initializer_range": 0.02,
    "intermediate_size": 5120,
    "layer_types": [
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention",
      "full_attention"
    ],
    "max_position_embeddings": 8192,
    "max_window_layers": 28,
    "model_type": "llada2_vl",
    "moe_intermediate_size": 512,
    "moe_layer_freq": [],
    "moe_layers": [
      1,
      2,
      3,
      4,
      5,
      6,
      7,
      8,
      9,
      10,
      11,
      12,
      13,
      14,
      15,
      16,
      17,
      18,
      19
    ],
    "moe_router_enable_expert_bias": true,
    "moe_router_score_function": null,
    "n_group": 8,
    "norm_head": false,
    "norm_softmax": false,
    "norm_topk_prob": true,
    "num_attention_heads": 16,
    "num_experts": 256,
    "num_experts_per_tok": 8,
    "num_hidden_layers": 20,
    "num_key_value_heads": 4,
    "num_shared_experts": 1,
    "output_dropout": 0.0,
    "output_router_logits": false,
    "pad_token_id": 156892,
    "partial_rotary_factor": 0.5,
    "qk_layernorm": null,
    "rms_norm_eps": 1e-06,
    "rope_theta": 600000,
    "rotary_dim": 64,
    "routed_scaling_factor": 2.5,
    "router_aux_loss_coef": 0.01,
    "router_dtype": "fp32",
    "router_num_group": -2,
    "router_topk_group": -2,
    "score_function": "sigmoid",
    "shared_expert_intermediate_size": -1,
    "sliding_window": null,
    "topk_group": 4,
    "torch_dtype": "bfloat16",
    "use_bias": false,
    "use_cache": false,
    "use_qkv_bias": false,
    "use_rmsnorm": true,
    "use_sliding_window": false,
    "using_split_qkv_in_self_attention": false,
    "vocab_size": 157188
  },
  "transformers_version": "4.52.3",
  "video_token_id": 151656111,
  "vision_config": {
    "depth": 32,
    "fullatt_block_indexes": [
      7,
      15,
      23,
      31
    ],
    "hidden_act": "silu",
    "hidden_size": 1280,
    "in_channels": 3,
    "initializer_range": 0.02,
    "intermediate_size": 3420,
    "model_type": "lladavision",
    "num_heads": 16,
    "out_hidden_size": 2048,
    "patch_size": 14,
    "spatial_merge_size": 2,
    "temporal_patch_size": 2,
    "tokens_per_second": 2,
    "torch_dtype": "bfloat16",
    "window_size": 112
  },
  "vision_end_token_id": 157186,
  "vision_start_token_id": 157185
}