{
    "hidden_size": 1024,
    "num_hidden_layers": 24,
    "layer_norm_eps": 1e-05,
    "num_attention_heads": 16,
    "hidden_act": "quick_gelu",
    "attention_dropout": 0.0,
    "intermediate_size": 4096,
    "initializer_range": 0.02,
    "image_size": 336,
    "num_channels": 3,
    "patch_size": 14,
    "projection_dim": 512,
    "model_type": "clip_vision_model"
}