config.json

{
  "architectures": [
    "Ernie4_5_VLMoeForConditionalGeneration"
  ],
  "hidden_act": "silu",
  "hidden_size": 8192,
  "intermediate_size": 28672,
  "max_position_embeddings": 131072,
  "num_attention_heads": 64,
  "num_key_value_heads": 8,
  "num_hidden_layers": 54,
  "model_type": "ernie4_5_moe_vl",
  "rms_norm_eps": 1e-05,
  "vocab_size": 103424,
  "rope_theta": 500000,
  "use_rmsnorm": true,
  "use_bias": false,
  "moe_layer_start_index": 3,
  "moe_capacity": [64,64,64],
  "moe_gate": "topk",
  "moe_k": 8,
  "moe_layer_interval": 1,
  "loss_subbatch_seqlen": 8192,
  "rope_3d": true,
  "freq_allocation": 20,
  "spatial_conv_size": 2,
  "temporal_conv_size": 2,
  "moe_use_aux_free": true,
  "moe_num_experts": [64, 64],
  "moe_intermediate_size": [3584, 1536],
  "dtype": "bfloat16",
  "tie_word_embeddings": 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,
    "attn_sep": true
  }
}