| # STFT arguments | |
| sample_rate: 16000 | |
| n_fft: 512 | |
| win_length: 32 | |
| hop_length: 16 | |
| mask_weight: 0.99 | |
| # Enhancement model args | |
| enhance_model: !new:speechbrain.lobes.models.EnhanceResnet.EnhanceResnet | |
| n_fft: !ref <n_fft> | |
| win_length: !ref <win_length> | |
| hop_length: !ref <hop_length> | |
| sample_rate: !ref <sample_rate> | |
| channel_counts: [128, 128, 256, 256, 512, 512] | |
| normalization: !name:speechbrain.nnet.normalization.BatchNorm2d | |
| activation: !new:torch.nn.GELU | |
| dense_count: 2 | |
| dense_nodes: 1024 | |
| dropout: 0.1 | |
| mask_weight: !ref <mask_weight> | |
| modules: | |
| enhance_model: !ref <enhance_model> | |
| pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer | |
| loadables: | |
| enhance_model: !ref <enhance_model> | |