| { | |
| "architectures": [ | |
| "MotionRVQ_VAE" | |
| ], | |
| "model_type": "motion_rvq_vae", | |
| "library_name": "pytorch", | |
| "torch_dtype": "float32", | |
| "in_channels": 263, | |
| "out_channels": 263, | |
| "latent_dim": 512, | |
| "encoder_hidden_channels": 512, | |
| "decoder_hidden_channels": 512, | |
| "rvq_num_levels": 4, | |
| "rvq_num_embeddings": 1024, | |
| "rvq_embedding_dim": 512, | |
| "rvq_commitment_cost": 0.25, | |
| "rvq_decay": 0.99, | |
| "rvq_epsilon": 1e-5, | |
| "downsample_stride": 4, | |
| "window_size": 100, | |
| "input_representation": "263-dim frame vectors", | |
| "task": "motion_reconstruction" | |
| } | |