| { | |
| "_class_name": "KandinskyS2VTransformer3DModel", | |
| "_diffusers_version": "0.40.0", | |
| "audio_inject_layers": [ | |
| 0 | |
| ], | |
| "axes_dims": [ | |
| 4, | |
| 6, | |
| 6 | |
| ], | |
| "enable_framepack": true, | |
| "ff_dim": 64, | |
| "frame_buckets": [ | |
| 1, | |
| 2, | |
| 4 | |
| ], | |
| "in_audio_dim": 16, | |
| "in_text_dim": 16, | |
| "in_text_dim2": 16, | |
| "in_visual_dim": 4, | |
| "index_rope_ref_image": 40, | |
| "model_dim": 32, | |
| "num_audio_blocks": 1, | |
| "num_text_blocks": 1, | |
| "num_visual_blocks": 2, | |
| "out_visual_dim": 4, | |
| "patch_proj1x_coef": [ | |
| 1, | |
| 1, | |
| 1 | |
| ], | |
| "patch_proj2x_coef": [ | |
| 2, | |
| 2, | |
| 2 | |
| ], | |
| "patch_proj4x_coef": [ | |
| 4, | |
| 4, | |
| 4 | |
| ], | |
| "patch_size": [ | |
| 1, | |
| 2, | |
| 2 | |
| ], | |
| "text_rope_max_pos": 128, | |
| "time_dim": 16 | |
| } | |