config.json

{
  "model_type": "clip",
  "img_encoder":{
    "backbone": "deit3_base_patch16_224_in21ft1k",
    "dim": 768,
    "pooling": "cls",
    "output_dim": 256,
    "backbone_type": "vit"
  },
  "text_encoder":{
    "backbone": "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
    "backbone_type": "bert",
    "unimodal_n_layers": 8,
    "dim": 384,
    "pooling": "mean",
    "context_dim": 768,
    "output_dim": 256,
    "head_one_neuron": true
  }
}