config.json
| 1 | { |
| 2 | "architectures": [ |
| 3 | "SiglipModel" |
| 4 | ], |
| 5 | "initializer_factor": 1.0, |
| 6 | "model_type": "siglip", |
| 7 | "text_config": { |
| 8 | "hidden_size": 1152, |
| 9 | "intermediate_size": 4304, |
| 10 | "model_type": "siglip_text_model", |
| 11 | "num_attention_heads": 16, |
| 12 | "num_hidden_layers": 27 |
| 13 | }, |
| 14 | "torch_dtype": "float32", |
| 15 | "transformers_version": "4.37.0.dev0", |
| 16 | "vision_config": { |
| 17 | "hidden_size": 1152, |
| 18 | "image_size": 384, |
| 19 | "intermediate_size": 4304, |
| 20 | "model_type": "siglip_vision_model", |
| 21 | "num_attention_heads": 16, |
| 22 | "num_hidden_layers": 27, |
| 23 | "patch_size": 14 |
| 24 | } |
| 25 | } |
| 26 | |