vision_tower/config.json

{
  "configs": {},
  "model_type": "hybrid_vision_tower",
  "num_vision_encoders": 2,
  "token_output_shape": [
    4,
    27,
    27
  ],
  "transformers_version": "4.44.0",
  "vision_towers": [
    "siglip-so400m-patch14-384",
    "internvideo2"
  ],
  "auto_map": {
    "AutoConfig": "configuration_hybrid.HybridTowerConfig"
  }
}