config.json

{
  "activation_dropout": 0.0,
  "activation_fn": "gelu",
  "architectures": [
    "Beit3ForImageTextRetrieval"
  ],
  "attention_dropout": 0.0,
  "deepnorm": false,
  "dropout": 0.0,
  "embed_dim": 768,
  "encoder_normalize_before": true,
  "hidden_size": 3072,
  "img_size": 384,
  "initializer_range": 0.02,
  "label_smoothing": 0.1,
  "layer_norm_eps": 1e-05,
  "layers": 12,
  "logit_scale_init_value": 2.65926,
  "max_source_positions": 1024,
  "model_type": "beit3",
  "normalize_before": true,
  "num_attention_heads": 12,
  "num_channels": 3,
  "patch_size": 16,
  "sub_layernorm": true,
  "torch_dtype": "float32",
  "transformers_version": "4.34.0.dev0",
  "vocab_size": 64010
}