| { | |
| "auto_map": { | |
| "AutoConfig": "configuration_skysensepp.SkySensePPConfig", | |
| "AutoModel": "modeling_skysensepp.SkySensePPModel" | |
| }, | |
| "architectures": [ | |
| "SkySensePPModel" | |
| ], | |
| "model_type": "skysensepp", | |
| "sources": [ | |
| "hr", | |
| "s2", | |
| "s1" | |
| ], | |
| "use_modal_vae": true, | |
| "calendar_time": 366, | |
| "hr_arch": "huge", | |
| "hr_img_size": 512, | |
| "hr_patch_size": 4, | |
| "hr_in_channels": 3, | |
| "hr_window_size": 8, | |
| "hr_drop_path_rate": 0.2, | |
| "hr_out_indices": [ | |
| 0, | |
| 1, | |
| 2, | |
| 3 | |
| ], | |
| "hr_use_abs_pos_embed": false, | |
| "hr_with_cp": true, | |
| "hr_pad_small_map": true, | |
| "s2_img_size": [ | |
| 16, | |
| 16 | |
| ], | |
| "s2_patch_size": 4, | |
| "s2_in_channels": 10, | |
| "s2_embed_dims": 1024, | |
| "s2_num_layers": 24, | |
| "s2_num_heads": 16, | |
| "s2_mlp_ratio": 4, | |
| "s2_out_indices": [ | |
| 5, | |
| 11, | |
| 17, | |
| 23 | |
| ], | |
| "s2_drop_path_rate": 0.3, | |
| "s1_img_size": [ | |
| 16, | |
| 16 | |
| ], | |
| "s1_patch_size": 4, | |
| "s1_in_channels": 2, | |
| "s1_embed_dims": 1024, | |
| "s1_num_layers": 24, | |
| "s1_num_heads": 16, | |
| "fusion_input_dims": 2816, | |
| "fusion_embed_dims": 1024, | |
| "fusion_num_layers": 24, | |
| "fusion_num_heads": 16, | |
| "fusion_with_cls_token": true, | |
| "fusion_output_cls_token": true, | |
| "decode_in_channels": [ | |
| 704, | |
| 704, | |
| 1408, | |
| 2816, | |
| 1024 | |
| ], | |
| "decode_channels": 512, | |
| "decode_num_classes": 65, | |
| "vocabulary_size": 64, | |
| "vae_subfolder": "modality_vae" | |
| } |