config.json

{
  "embed_dim": 512,
  "text_cfg": {
    "context_length": 77,
    "heads": 8,
    "layers": 12,
    "unc_layers": 2,
    "vocab_size": 49408,
    "width": 512
  },
  "vision_cfg": {
    "image_size": 224,
    "layers": 12,
    "patch_size": 16,
    "unc_layers": 2,
    "width": 768
  }
}