| fg_dir: !!python/object/apply:pathlib.PosixPath | |
| - data | |
| - fg | |
| bg_dir: !!python/object/apply:pathlib.PosixPath | |
| - data | |
| - bg | |
| rir_dir: !!python/object/apply:pathlib.PosixPath | |
| - data | |
| - rir | |
| load_fg_only: false | |
| wav_rate: 44100 | |
| n_fft: 2048 | |
| win_size: 2048 | |
| hop_size: 420 | |
| num_mels: 128 | |
| stft_magnitude_min: 0.0001 | |
| preemphasis: 0.97 | |
| mix_alpha_range: | |
| - 0.2 | |
| - 0.8 | |
| nj: 64 | |
| training_seconds: 3.0 | |
| batch_size_per_gpu: 32 | |
| min_lr: 1.0e-05 | |
| max_lr: 0.0001 | |
| warmup_steps: 1000 | |
| max_steps: 1000000 | |
| gradient_clipping: 1.0 | |
| cfm_solver_method: midpoint | |
| cfm_solver_nfe: 64 | |
| cfm_time_mapping_divisor: 4 | |
| univnet_nc: 96 | |
| lcfm_latent_dim: 64 | |
| lcfm_training_mode: cfm | |
| lcfm_z_scale: 6 | |
| vocoder_extra_dim: 32 | |
| gan_training_start_step: null | |
| praat_augment_prob: 0.2 | |