{
  "architectures": [
    "Ernie4_5_VLMoeForConditionalGeneration"
  ],
  "dtype": "bfloat16",
  "hidden_act": "silu",
  "hidden_size": 2560,
  "intermediate_size": 12288,
  "model_type": "ernie4_5_moe_vl",
  "moe_capacity": [128, 128, 128],
  "moe_gate": "topk",
  "moe_intermediate_size": [1536, 512],
  "moe_k": 6,
  "moe_layer_end_index": [29, 28],
  "moe_layer_interval": 1,
  "moe_layer_start_index": [1, 1],
  "moe_multimodal_dispatch_use_allgather": "v2-alltoall-unpad-text",
  "moe_num_experts": [64, 64],
  "moe_num_shared_experts": 2,
  "moe_use_aux_free": true,
  "num_attention_heads": 20,
  "num_hidden_layers": 28,
  "num_key_value_heads": 4,
  "pixel_hidden_size": 1280,
  "rms_norm_eps": 1e-05,
  "rope_3d": true,
  "rope_theta": 500000,
  "spatial_conv_size": 2,
  "temporal_conv_size": 2,
  "vocab_size": 103424,
  "tie_word_embeddings": true,
  "use_cache": true,
  "use_rmsnorm": true,
  "use_bias": false,
  "vision_config": {
    "attn_implementation": "eager",
    "depth": 32,
    "embed_dim": 1280,
    "hidden_act": "quick_gelu",
    "hidden_size": 1280,
    "in_channels": 3,
    "in_chans": 3,
    "mlp_ratio": 4,
    "num_heads": 16,
    "patch_size": 14,
    "spatial_merge_size": 2,
    "spatial_patch_size": 14,
    "vit_first_fwd_bsz": 128,
    "attn_sep": true
  }
}