{
  "architectures": [
    "TwoTowerBridgeTransformer"
  ],
  "attention_implementation": "sdpa",
  "attention_probability_dropout_probability": 0.1,
  "dropout_probability": 0.1,
  "embedding_size": 512,
  "hidden_activation_function": "gelu",
  "hidden_size": 768,
  "init_layer_norm_from_vision_encoder": false,
  "initializer_factor": 1.0,
  "intermediate_size": 3072,
  "layer_norm_eps": 1e-05,
  "link_tower_type": "add",
  "num_attention_heads": 12,
  "num_hidden_layers": 6,
  "num_intermediate_layers": 1,
  "share_cross_modal_transformer_layers": true,
  "share_link_tower_layers": false,
  "text_configuration": {
    "_attn_implementation_autoset": true,
    "model_type": ""
  },
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "transformers_version": "4.48.3",
  "vision_configuration": {
    "_attn_implementation_autoset": true,
    "model_type": ""
  }
}