| { | |
| "architectures": [ | |
| "BlipForConditionalGeneration" | |
| ], | |
| "model_type": "blip", | |
| "text_config": { | |
| "hidden_size": 768, | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 12 | |
| }, | |
| "vision_config": { | |
| "hidden_size": 1024, | |
| "image_size": 384, | |
| "patch_size": 16 | |
| } | |
| } |
| { | |
| "architectures": [ | |
| "BlipForConditionalGeneration" | |
| ], | |
| "model_type": "blip", | |
| "text_config": { | |
| "hidden_size": 768, | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 12 | |
| }, | |
| "vision_config": { | |
| "hidden_size": 1024, | |
| "image_size": 384, | |
| "patch_size": 16 | |
| } | |
| } |