| audio: | |
| chunk_size: 260096 | |
| dim_f: 4096 | |
| dim_t: 128 | |
| hop_length: 2048 | |
| n_fft: 12288 | |
| num_channels: 2 | |
| sample_rate: 44100 | |
| model: | |
| act: gelu | |
| bottleneck_factor: 4 | |
| growth: 64 | |
| norm: InstanceNorm | |
| num_blocks_per_scale: 2 | |
| num_channels: 128 | |
| num_scales: 5 | |
| num_subbands: 4 | |
| scale: | |
| - 2 | |
| - 2 | |
| training: | |
| batch_size: 8 | |
| grad_clip: 0 | |
| instruments: | |
| - Vocals | |
| - Drums | |
| - Bass | |
| - Other | |
| lr: 5.0e-05 | |
| target_instrument: Vocals | |
| inference: | |
| batch_size: 1 | |
| dim_t: 1024 | |
| num_overlap: 8 |