-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #4387 from LiChenda/dynamic_mixing
Add dynamic mixing in the speech separation task.
- Loading branch information
Showing
8 changed files
with
356 additions
and
14 deletions.
There are no files selected for viewing
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
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
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
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 +1,37 @@ | ||
|
||
encoder: stft | ||
encoder_conf: | ||
n_fft: 512 | ||
hop_length: 128 | ||
decoder: stft | ||
decoder_conf: | ||
n_fft: 512 | ||
hop_length: 128 | ||
separator: rnn | ||
separator_conf: | ||
rnn_type: blstm | ||
num_spk: 2 | ||
nonlinear: relu | ||
layer: 1 | ||
unit: 128 | ||
dropout: 0.2 | ||
|
||
# dynamic_mixing related | ||
# dynamic_mixing_gain_db: | ||
# The maximum random gain (in dB) for each source before the mixing. | ||
# The gain (in dB) of each source is unifromly sampled in | ||
# [-dynamic_mixing_gain_db, dynamic_mixing_gain_db] | ||
dynamic_mixing: True | ||
dynamic_mixing_gain_db: 2.0 | ||
|
||
criterions: | ||
# The first criterion | ||
- name: mse | ||
conf: | ||
compute_on_mask: True | ||
mask_type: PSM^2 | ||
# the wrapper for the current criterion | ||
# for single-talker case, we simplely use fixed_order wrapper | ||
wrapper: fixed_order | ||
wrapper_conf: | ||
weight: 1.0 |
77 changes: 77 additions & 0 deletions
77
egs2/wsj0_2mix/enh1/conf/tuning/train_enh_skim_tasnet_noncausal_dm.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 +1,77 @@ | ||
init: xavier_uniform | ||
max_epoch: 150 | ||
batch_type: folded | ||
# When dynamic mixing is enabled, the actual batch_size will | ||
# be (batch_size / num_spk) | ||
batch_size: 8 | ||
iterator_type: chunk | ||
chunk_length: 16000 | ||
num_workers: 2 | ||
optim: adamw | ||
optim_conf: | ||
lr: 1.0e-03 | ||
eps: 1.0e-06 | ||
weight_decay: 0 | ||
patience: 20 | ||
grad_clip: 5.0 | ||
val_scheduler_criterion: | ||
- valid | ||
- loss | ||
best_model_criterion: | ||
- - valid | ||
- si_snr | ||
- max | ||
- - valid | ||
- loss | ||
- min | ||
keep_nbest_models: 10 | ||
|
||
scheduler: steplr | ||
scheduler_conf: | ||
step_size: 2 | ||
gamma: 0.97 | ||
|
||
# dynamic_mixing related | ||
# dynamic_mixing_gain_db: | ||
# The maximum random gain (in dB) for each source before the mixing. | ||
# The gain (in dB) of each source is unifromly sampled in | ||
# [-dynamic_mixing_gain_db, dynamic_mixing_gain_db] | ||
dynamic_mixing: True | ||
dynamic_mixing_gain_db: 2.0 | ||
|
||
encoder: conv | ||
encoder_conf: | ||
channel: 64 | ||
kernel_size: 2 | ||
stride: 1 | ||
decoder: conv | ||
decoder_conf: | ||
channel: 64 | ||
kernel_size: 2 | ||
stride: 1 | ||
separator: skim | ||
separator_conf: | ||
causal: False | ||
num_spk: 2 | ||
layer: 6 | ||
nonlinear: relu | ||
unit: 128 | ||
segment_size: 250 | ||
dropout: 0.05 | ||
mem_type: hc | ||
seg_overlap: True | ||
|
||
# A list for criterions | ||
# The overlall loss in the multi-task learning will be: | ||
# loss = weight_1 * loss_1 + ... + weight_N * loss_N | ||
# The default `weight` for each sub-loss is 1.0 | ||
criterions: | ||
# The first criterion | ||
- name: si_snr | ||
conf: | ||
eps: 1.0e-6 | ||
wrapper: pit | ||
wrapper_conf: | ||
weight: 1.0 | ||
independent_perm: True | ||
|
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
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
Oops, something went wrong.