forked from espnet/espnet
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
update egs2/chime4/enh_asr1/README.md and related enh1, asr1 configs.
- Loading branch information
1 parent
2b66331
commit de81cf9
Showing
3 changed files
with
212 additions
and
18 deletions.
There are no files selected for viewing
90 changes: 90 additions & 0 deletions
90
...r1/conf/tuning/train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k.yaml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,90 @@ | ||
| # minibatch related | ||
| batch_type: folded | ||
| batch_size: 32 | ||
| accum_grad: 1 | ||
| grad_clip: 5 | ||
| max_epoch: 50 | ||
| patience: none | ||
| # The initialization method for model parameters | ||
| init: xavier_uniform | ||
| val_scheduler_criterion: | ||
| - valid | ||
| - loss | ||
| best_model_criterion: | ||
| - - valid | ||
| - acc | ||
| - max | ||
| keep_nbest_models: 10 | ||
| unused_parameters: true | ||
| # SSL-based frontend is fixed during training for training efficiency, | ||
| # however, the gradients are backprogated through frontend to the enhancement. | ||
| freeze_param: [ | ||
| "frontend.upstream" | ||
| ] | ||
|
|
||
| # network architecture | ||
| frontend: s3prl | ||
| frontend_conf: | ||
| frontend_conf: | ||
| upstream: wavlm_large # Note: If the upstream is changed, please change the input_size in the preencoder. | ||
| download_dir: ./hub | ||
| multilayer_feature: True | ||
|
|
||
| preencoder: linear | ||
| preencoder_conf: | ||
| input_size: 1024 # Note: If the upstream is changed, please change this value accordingly. | ||
| output_size: 128 | ||
|
|
||
| # encoder related | ||
| encoder: transformer | ||
| encoder_conf: | ||
| output_size: 256 | ||
| attention_heads: 4 | ||
| linear_units: 2048 | ||
| num_blocks: 12 | ||
| dropout_rate: 0.1 | ||
| attention_dropout_rate: 0.0 | ||
| input_layer: conv2d2 | ||
| normalize_before: true | ||
|
|
||
| # decoder related | ||
| decoder: transformer | ||
| decoder_conf: | ||
| input_layer: embed | ||
| attention_heads: 4 | ||
| linear_units: 2048 | ||
| num_blocks: 6 | ||
| dropout_rate: 0.1 | ||
| positional_dropout_rate: 0.0 | ||
| self_attention_dropout_rate: 0.0 | ||
| src_attention_dropout_rate: 0.0 | ||
|
|
||
| model_conf: | ||
| ctc_weight: 0.3 | ||
| lsm_weight: 0.1 | ||
| length_normalized_loss: false | ||
| extract_feats_in_collect_stats: false | ||
|
|
||
| optim: adam | ||
| optim_conf: | ||
| lr: 0.001 | ||
| scheduler: warmuplr | ||
| scheduler_conf: | ||
| warmup_steps: 20000 | ||
|
|
||
| specaug: specaug | ||
| specaug_conf: | ||
| apply_time_warp: true | ||
| time_warp_window: 5 | ||
| time_warp_mode: bicubic | ||
| apply_freq_mask: true | ||
| freq_mask_width_range: | ||
| - 0 | ||
| - 100 | ||
| num_freq_mask: 4 | ||
| apply_time_mask: true | ||
| time_mask_width_range: | ||
| - 0 | ||
| - 40 | ||
| num_time_mask: 2 | ||
|
|
64 changes: 64 additions & 0 deletions
64
egs2/chime4/enh1/conf/tuning/train_enh_convtasnet_small.yaml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,64 @@ | ||
| optim: adam | ||
| init: xavier_uniform | ||
| max_epoch: 100 | ||
| batch_type: folded | ||
| batch_size: 32 | ||
| iterator_type: chunk | ||
| chunk_length: 32000 | ||
| num_workers: 4 | ||
| optim_conf: | ||
| lr: 1.0e-03 | ||
| eps: 1.0e-08 | ||
| weight_decay: 1.0e-05 | ||
| patience: 4 | ||
| val_scheduler_criterion: | ||
| - valid | ||
| - loss | ||
| best_model_criterion: | ||
| - - valid | ||
| - si_snr | ||
| - max | ||
| - - valid | ||
| - loss | ||
| - min | ||
| keep_nbest_models: 1 | ||
| scheduler: reducelronplateau | ||
| scheduler_conf: | ||
| mode: min | ||
| factor: 0.5 | ||
| patience: 3 | ||
| model_conf: | ||
| loss_type: si_snr | ||
| encoder: conv | ||
| encoder_conf: | ||
| channel: 256 | ||
| kernel_size: 40 | ||
| stride: 20 | ||
| decoder: conv | ||
| decoder_conf: | ||
| channel: 256 | ||
| kernel_size: 40 | ||
| stride: 20 | ||
| separator: tcn | ||
| separator_conf: | ||
| num_spk: 1 | ||
| layer: 4 | ||
| stack: 2 | ||
| bottleneck_dim: 256 | ||
| hidden_dim: 512 | ||
| kernel: 3 | ||
| causal: False | ||
| norm_type: "gLN" | ||
| nonlinear: relu | ||
| criterions: | ||
| # The first criterion | ||
| - name: si_snr | ||
| conf: | ||
| eps: 1e-7 | ||
| # the wrapper for the current criterion | ||
| # for single-talker case, we simplely use fixed_order wrapper | ||
| wrapper: | ||
| - type: fixed_order | ||
| wrapper_conf: | ||
| weight: 1.0 | ||
|
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters