| { | |
| "architectures": [ | |
| "SiglipModel" | |
| ], | |
| "model_type": "siglip", | |
| "text_config": { | |
| "architectures": [ | |
| "SiglipTextModel" | |
| ], | |
| "attention_dropout": 0.0, | |
| "dropout": 0.0, | |
| "hidden_act": "gelu", | |
| "hidden_size": 1152, | |
| "initializer_factor": 1.0, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 4608, | |
| "layer_norm_eps": 1e-06, | |
| "max_position_embeddings": 64, | |
| "model_type": "siglip_text_model", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 32, | |
| "pad_token_id": 0, | |
| "vocab_size": 32000 | |
| }, | |
| "vision_config": { | |
| "architectures": [ | |
| "SiglipVisionModel" | |
| ], | |
| "attention_dropout": 0.0, | |
| "dropout": 0.0, | |
| "hidden_act": "gelu", | |
| "hidden_size": 1152, | |
| "image_size": 384, | |
| "initializer_factor": 1.0, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 4608, | |
| "layer_norm_eps": 1e-06, | |
| "model_type": "siglip_vision_model", | |
| "num_attention_heads": 16, | |
| "num_channels": 3, | |
| "num_hidden_layers": 32, | |
| "patch_size": 16 | |
| }, | |
| "vision_dim": 1152, | |
| "text_dim": 384, | |
| "embedding_dim": 512, | |
| "note": "Includes text encoder projections and gating mechanism" | |
| } |