| model_name: "silero_vad" | |
| # spec | |
| sample_rate: 8000 | |
| nfft: 512 | |
| win_size: 240 | |
| hop_size: 80 | |
| win_type: hann | |
| # model | |
| encoder_in_channels: 64 | |
| encoder_hidden_channels: 128 | |
| encoder_out_channels: 64 | |
| encoder_kernel_size: 3 | |
| encoder_num_layers: 3 | |
| decoder_hidden_size: 64 | |
| decoder_num_layers: 2 | |
| # lsnr | |
| n_frame: 3 | |
| min_local_snr_db: -15 | |
| max_local_snr_db: 30 | |
| norm_tau: 1. | |
| # data | |
| min_snr_db: -10 | |
| max_snr_db: 20 | |
| # train | |
| lr: 0.001 | |
| lr_scheduler: "CosineAnnealingLR" | |
| lr_scheduler_kwargs: | |
| T_max: 250000 | |
| eta_min: 0.0001 | |
| max_epochs: 100 | |
| clip_grad_norm: 10.0 | |
| seed: 1234 | |
| num_workers: 4 | |
| batch_size: 128 | |
| eval_steps: 25000 | |