| # params.yaml | |
| data: | |
| target_sr: 16000 | |
| window_sec: 4.0 | |
| stride_sec: 2.0 | |
| min_rms: 0.001 | |
| augment: | |
| pitch_steps_low: 4.0 | |
| pitch_steps_high: 8.0 | |
| model: | |
| wavlm_name: "microsoft/wavlm-large" | |
| acoustic_dim: 256 | |
| lpe_dim: 32 | |
| hidden_dim: 128 | |
| dropout: 0.3 | |
| train: | |
| batch_size_stage1: 16 | |
| batch_size_stage2: 16 | |
| epochs_stage1: 20 | |
| epochs_stage2: 30 | |
| lr_wavlm_stage1: 0.00001 | |
| lr_wavlm_stage2: 0.000002 | |
| lr_heads_stage1: 0.0001 | |
| lr_heads_stage2: 0.00005 | |
| pos_weight_sld: 3.0 | |
| pos_weight_sld_stage2: 0.25 # neg/pos = 706/2841 | |
| pos_weight_od: 2.0 | |