| { |
| "architectures": [ |
| "CanterModel" |
| ], |
| "canter": { |
| "adaln_rank": 256, |
| "cross_head_dim": 128, |
| "cross_heads": 16, |
| "depth": 30, |
| "image_heads": 16, |
| "latent_channels": 128, |
| "mlp_ratio": 4, |
| "prefix_depth": 3, |
| "rope_base": 10000.0, |
| "suffix_depth": 3, |
| "text_backbone_width": 960, |
| "text_heads": 8, |
| "text_max_length": 7999, |
| "text_refine_depth": 4, |
| "text_rope_theta": 160000.0, |
| "text_width": 1024, |
| "time_frequency_width": 256, |
| "time_scale": 1000.0, |
| "width": 2048 |
| }, |
| "model_type": "canter", |
| "torch_dtype": "bfloat16", |
| "weights": "ema" |
| } |
|
|
