{
  "export_model_type": "clip",
  "initializer_factor": 1.0,
  "logit_scale_init_value": 2.6592,
  "model_cfg": {
    "custom_text": true,
    "embed_dim": 512,
    "text_cfg": {
      "context_length": 77,
      "heads": 8,
      "layers": 12,
      "no_causal_mask": true,
      "vocab_size": 49408,
      "width": 512
    },
    "vision_cfg": {
      "image_size": 256,
      "timm_drop": 0.0,
      "timm_drop_path": 0.0,
      "timm_model_name": "fastvit_mci1",
      "timm_model_pretrained": false,
      "timm_pool": "avg",
      "timm_proj": null
    }
  },
  "model_type": "clip",
  "preprocess_cfg": {
    "interpolation": "bilinear",
    "mean": [
      0.0,
      0.0,
      0.0
    ],
    "resize_mode": "shortest",
    "std": [
      1.0,
      1.0,
      1.0
    ]
  },
  "projection_dim": 512,
  "text_config": {
    "context_length": 77,
    "heads": 8,
    "layers": 12,
    "model_type": "clip_text_model",
    "no_causal_mask": true,
    "width": 512
  },
  "transformers_version": "4.41.2",
  "vision_config": {
    "image_size": 256,
    "model_type": "clip_vision_model",
    "timm_drop": 0.0,
    "timm_drop_path": 0.0,
    "timm_model_name": "fastvit_mci1",
    "timm_model_pretrained": false,
    "timm_pool": "avg",
    "timm_proj": null
  }
}