config.js

{
  "architectures": [
    "SCOLD"
  ],
  "model_type": "clip-base",
  "image_encoder": "swin_base_patch4_window7_224",
  "text_encoder": "roberta-base",
  "embedding_dim": 512,
  "t_init": 0.07,
  "b_init": 0.0
}