config.json

{
  "architectures": [
    "LlavaForConditionalGeneration"
  ],
  "dtype": "float32",
  "ignore_index": -100,
  "image_seq_length": 196,
  "image_token_index": 32768,
  "model_type": "llava",
  "multimodal_projector_bias": true,
  "projector_hidden_act": "gelu",
  "text_config": {
    "_name_or_path": "SmallDoge/Doge-20M",
    "architectures": [
      "DogeForCausalLM"
    ],
    "attention_bias": false,
    "attention_dropout": 0.0,
    "dtype": "float32",
    "hidden_act": "silu",
    "hidden_dropout": 0.0,
    "hidden_size": 256,
    "initializer_range": 0.02,
    "intermediate_size": 512,
    "is_moe": false,
    "keep_window_size": 2048,
    "max_position_embeddings": 2048,
    "mlp_bias": false,
    "model_type": "doge",
    "norm_topk_prob": false,
    "num_attention_heads": 2,
    "num_experts": 16384,
    "num_experts_per_tok": 64,
    "num_hidden_layers": 8,
    "num_key_value_heads": 1,
    "output_router_logits": false,
    "rms_norm_eps": 1e-06,
    "rope_scaling": null,
    "rope_theta": 10000.0,
    "router_aux_loss_coef": 0.001,
    "sliding_window": null,
    "tie_word_embeddings": true,
    "use_cache": true,
    "vocab_size": 32769
  },
  "transformers_version": "4.57.3",
  "vision_config": {
    "_name_or_path": "google/siglip2-base-patch16-224",
    "attention_dropout": 0.0,
    "dtype": "float32",
    "hidden_act": "gelu_pytorch_tanh",
    "hidden_size": 768,
    "image_size": 224,
    "intermediate_size": 3072,
    "layer_norm_eps": 1e-06,
    "model_type": "siglip_vision_model",
    "num_attention_heads": 12,
    "num_channels": 3,
    "num_hidden_layers": 12,
    "output_hidden_states": true,
    "patch_size": 16
  },
  "vision_feature_layer": -1,
  "vision_feature_select_strategy": "full"
}