config.json

{
  "_name_or_path": "BridgeTower/bridgetower-large-itm-mlm-itc",
  "architectures": [
    "BridgeTowerForContrastiveLearning"
  ],
  "contrastive_hidden_size": 512,
  "drop_rate": 0.1,
  "head_hidden_scale": 2,
  "hidden_act": "gelu",
  "hidden_size": 1024,
  "init_layernorm_from_vision_encoder": false,
  "initializer_factor": 1,
  "layer_norm_eps": 1e-05,
  "link_tower_type": "add",
  "logit_scale_init_value": 2.6592,
  "model_type": "bridgetower",
  "num_attention_heads": 16,
  "num_hidden_layers": 6,
  "share_cross_modal_transformer_layers": true,
  "share_link_tower_layers": false,
  "text_config": {
    "architectures": [
      "BridgeTowerTextModel"
    ],
    "classifier_dropout": null,
    "hidden_size": 1024,
    "initializer_range": 0.02,
    "intermediate_size": 4096,
    "model_type": "bridgetower_text_model",
    "num_attention_heads": 16,
    "num_hidden_layers": 24
  },
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "transformers_version": "4.33.1",
  "vision_config": {
    "architectures": [
      "BridgeTowerVisionModel"
    ],
    "hidden_size": 1024,
    "image_size": 294,
    "model_type": "bridgetower_vision_model",
    "num_hidden_layers": 24,
    "patch_size": 14
  }
}