config.json
| 1 | { |
| 2 | "initializer_factor": 1.0, |
| 3 | "model_type": "siglip", |
| 4 | "text_config": { |
| 5 | "hidden_size": 1152, |
| 6 | "intermediate_size": 4304, |
| 7 | "model_type": "siglip_text_model", |
| 8 | "num_attention_heads": 16, |
| 9 | "num_hidden_layers": 27, |
| 10 | "projection_size": 1536, |
| 11 | "vocab_size": 256000 |
| 12 | }, |
| 13 | "transformers_version": "4.49.0.dev0", |
| 14 | "vision_config": { |
| 15 | "hidden_size": 1536, |
| 16 | "image_size": 384, |
| 17 | "intermediate_size": 6144, |
| 18 | "model_type": "siglip_vision_model", |
| 19 | "num_attention_heads": 16, |
| 20 | "num_hidden_layers": 40 |
| 21 | } |
| 22 | } |
| 23 | |