| { | |
| "_vocab_size": 257152, | |
| "architectures": [ | |
| "PaliGemmaForConditionalGeneration" | |
| ], | |
| "hidden_size": 2048, | |
| "image_token_index": 256000, | |
| "model_type": "paligemma", | |
| "projection_dim": 8, | |
| "text_config": { | |
| "hidden_size": 8, | |
| "intermediate_size": 32, | |
| "model_type": "gemma", | |
| "num_attention_heads": 4, | |
| "num_hidden_layers": 2, | |
| "num_image_tokens": 196, | |
| "num_key_value_heads": 2, | |
| "vocab_size": 258522 | |
| }, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.47.0.dev0", | |
| "vision_config": { | |
| "hidden_size": 8, | |
| "intermediate_size": 32, | |
| "model_type": "siglip_vision_model", | |
| "num_attention_heads": 4, | |
| "num_hidden_layers": 2, | |
| "patch_size": 16, | |
| "projection_dim": 8 | |
| } | |
| } | |