Update model
Browse files- README.md +266 -0
- exp/enh_stats_8k/train/feats_stats.npz +0 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/146epoch.pth +3 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/RESULTS.md +20 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/config.yaml +164 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/images/backward_time.png +0 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/images/forward_time.png +0 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/images/gpu_max_cached_mem_GB.png +0 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/images/iter_time.png +0 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/images/loss.png +0 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/images/optim0_lr0.png +0 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/images/optim_step_time.png +0 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/images/si_snr_loss.png +0 -0
- exp/enh_train_enh_skim_tasnet_noncausal_raw/images/train_time.png +0 -0
- meta.yaml +8 -0
README.md
ADDED
@@ -0,0 +1,266 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
tags:
|
3 |
+
- espnet
|
4 |
+
- audio
|
5 |
+
- audio-to-audio
|
6 |
+
language: en
|
7 |
+
datasets:
|
8 |
+
- wsj0_2mix
|
9 |
+
license: cc-by-4.0
|
10 |
+
---
|
11 |
+
|
12 |
+
## ESPnet2 ENH model
|
13 |
+
|
14 |
+
### `lichenda/wsj0_2mix_skim_noncausal`
|
15 |
+
|
16 |
+
This model was trained by LiChenda using wsj0_2mix recipe in [espnet](https://github.com/espnet/espnet/).
|
17 |
+
|
18 |
+
### Demo: How to use in ESPnet2
|
19 |
+
|
20 |
+
```bash
|
21 |
+
cd espnet
|
22 |
+
git checkout ac3c10cfe4faf82c0bb30f8b32d9e8692363e0a9
|
23 |
+
pip install -e .
|
24 |
+
cd egs2/wsj0_2mix/enh1
|
25 |
+
./run.sh --skip_data_prep false --skip_train true --download_model lichenda/wsj0_2mix_skim_noncausal
|
26 |
+
```
|
27 |
+
|
28 |
+
<!-- Generated by ./scripts/utils/show_enh_score.sh -->
|
29 |
+
# RESULTS
|
30 |
+
## Environments
|
31 |
+
- date: `Wed Feb 23 16:42:06 CST 2022`
|
32 |
+
- python version: `3.7.11 (default, Jul 27 2021, 14:32:16) [GCC 7.5.0]`
|
33 |
+
- espnet version: `espnet 0.10.7a1`
|
34 |
+
- pytorch version: `pytorch 1.8.1`
|
35 |
+
- Git hash: `ac3c10cfe4faf82c0bb30f8b32d9e8692363e0a9`
|
36 |
+
- Commit date: `Fri Feb 11 16:22:52 2022 +0800`
|
37 |
+
|
38 |
+
|
39 |
+
## ..
|
40 |
+
|
41 |
+
config: conf/tuning/train_enh_skim_tasnet_noncausal.yaml
|
42 |
+
|
43 |
+
|dataset|STOI|SAR|SDR|SIR|
|
44 |
+
|---|---|---|---|---|
|
45 |
+
|enhanced_cv_min_8k|0.96|19.17|18.70|29.56|
|
46 |
+
|enhanced_tt_min_8k|0.97|18.96|18.45|29.31|
|
47 |
+
|
48 |
+
## ENH config
|
49 |
+
|
50 |
+
<details><summary>expand</summary>
|
51 |
+
|
52 |
+
```
|
53 |
+
config: conf/tuning/train_enh_skim_tasnet_noncausal.yaml
|
54 |
+
print_config: false
|
55 |
+
log_level: INFO
|
56 |
+
dry_run: false
|
57 |
+
iterator_type: chunk
|
58 |
+
output_dir: exp/enh_train_enh_skim_tasnet_noncausal_raw
|
59 |
+
ngpu: 1
|
60 |
+
seed: 0
|
61 |
+
num_workers: 4
|
62 |
+
num_att_plot: 3
|
63 |
+
dist_backend: nccl
|
64 |
+
dist_init_method: env://
|
65 |
+
dist_world_size: null
|
66 |
+
dist_rank: null
|
67 |
+
local_rank: 0
|
68 |
+
dist_master_addr: null
|
69 |
+
dist_master_port: null
|
70 |
+
dist_launcher: null
|
71 |
+
multiprocessing_distributed: false
|
72 |
+
unused_parameters: false
|
73 |
+
sharded_ddp: false
|
74 |
+
cudnn_enabled: true
|
75 |
+
cudnn_benchmark: false
|
76 |
+
cudnn_deterministic: true
|
77 |
+
collect_stats: false
|
78 |
+
write_collected_feats: false
|
79 |
+
max_epoch: 150
|
80 |
+
patience: 20
|
81 |
+
val_scheduler_criterion:
|
82 |
+
- valid
|
83 |
+
- loss
|
84 |
+
early_stopping_criterion:
|
85 |
+
- valid
|
86 |
+
- loss
|
87 |
+
- min
|
88 |
+
best_model_criterion:
|
89 |
+
- - valid
|
90 |
+
- si_snr
|
91 |
+
- max
|
92 |
+
- - valid
|
93 |
+
- loss
|
94 |
+
- min
|
95 |
+
keep_nbest_models: 1
|
96 |
+
nbest_averaging_interval: 0
|
97 |
+
grad_clip: 5.0
|
98 |
+
grad_clip_type: 2.0
|
99 |
+
grad_noise: false
|
100 |
+
accum_grad: 1
|
101 |
+
no_forward_run: false
|
102 |
+
resume: true
|
103 |
+
train_dtype: float32
|
104 |
+
use_amp: false
|
105 |
+
log_interval: null
|
106 |
+
use_matplotlib: true
|
107 |
+
use_tensorboard: true
|
108 |
+
use_wandb: false
|
109 |
+
wandb_project: null
|
110 |
+
wandb_id: null
|
111 |
+
wandb_entity: null
|
112 |
+
wandb_name: null
|
113 |
+
wandb_model_log_interval: -1
|
114 |
+
detect_anomaly: false
|
115 |
+
pretrain_path: null
|
116 |
+
init_param: []
|
117 |
+
ignore_init_mismatch: false
|
118 |
+
freeze_param: []
|
119 |
+
num_iters_per_epoch: null
|
120 |
+
batch_size: 8
|
121 |
+
valid_batch_size: null
|
122 |
+
batch_bins: 1000000
|
123 |
+
valid_batch_bins: null
|
124 |
+
train_shape_file:
|
125 |
+
- exp/enh_stats_8k/train/speech_mix_shape
|
126 |
+
- exp/enh_stats_8k/train/speech_ref1_shape
|
127 |
+
- exp/enh_stats_8k/train/speech_ref2_shape
|
128 |
+
valid_shape_file:
|
129 |
+
- exp/enh_stats_8k/valid/speech_mix_shape
|
130 |
+
- exp/enh_stats_8k/valid/speech_ref1_shape
|
131 |
+
- exp/enh_stats_8k/valid/speech_ref2_shape
|
132 |
+
batch_type: folded
|
133 |
+
valid_batch_type: null
|
134 |
+
fold_length:
|
135 |
+
- 80000
|
136 |
+
- 80000
|
137 |
+
- 80000
|
138 |
+
sort_in_batch: descending
|
139 |
+
sort_batch: descending
|
140 |
+
multiple_iterator: false
|
141 |
+
chunk_length: 16000
|
142 |
+
chunk_shift_ratio: 0.5
|
143 |
+
num_cache_chunks: 1024
|
144 |
+
train_data_path_and_name_and_type:
|
145 |
+
- - dump/raw/tr_min_8k/wav.scp
|
146 |
+
- speech_mix
|
147 |
+
- sound
|
148 |
+
- - dump/raw/tr_min_8k/spk1.scp
|
149 |
+
- speech_ref1
|
150 |
+
- sound
|
151 |
+
- - dump/raw/tr_min_8k/spk2.scp
|
152 |
+
- speech_ref2
|
153 |
+
- sound
|
154 |
+
valid_data_path_and_name_and_type:
|
155 |
+
- - dump/raw/cv_min_8k/wav.scp
|
156 |
+
- speech_mix
|
157 |
+
- sound
|
158 |
+
- - dump/raw/cv_min_8k/spk1.scp
|
159 |
+
- speech_ref1
|
160 |
+
- sound
|
161 |
+
- - dump/raw/cv_min_8k/spk2.scp
|
162 |
+
- speech_ref2
|
163 |
+
- sound
|
164 |
+
allow_variable_data_keys: false
|
165 |
+
max_cache_size: 0.0
|
166 |
+
max_cache_fd: 32
|
167 |
+
valid_max_cache_size: null
|
168 |
+
optim: adam
|
169 |
+
optim_conf:
|
170 |
+
lr: 0.001
|
171 |
+
eps: 1.0e-08
|
172 |
+
weight_decay: 0
|
173 |
+
scheduler: reducelronplateau
|
174 |
+
scheduler_conf:
|
175 |
+
mode: min
|
176 |
+
factor: 0.7
|
177 |
+
patience: 1
|
178 |
+
init: xavier_uniform
|
179 |
+
model_conf:
|
180 |
+
stft_consistency: false
|
181 |
+
loss_type: mask_mse
|
182 |
+
mask_type: null
|
183 |
+
criterions:
|
184 |
+
- name: si_snr
|
185 |
+
conf:
|
186 |
+
eps: 1.0e-07
|
187 |
+
wrapper: pit
|
188 |
+
wrapper_conf:
|
189 |
+
weight: 1.0
|
190 |
+
independent_perm: true
|
191 |
+
use_preprocessor: false
|
192 |
+
encoder: conv
|
193 |
+
encoder_conf:
|
194 |
+
channel: 64
|
195 |
+
kernel_size: 2
|
196 |
+
stride: 1
|
197 |
+
separator: skim
|
198 |
+
separator_conf:
|
199 |
+
causal: false
|
200 |
+
num_spk: 2
|
201 |
+
layer: 6
|
202 |
+
nonlinear: relu
|
203 |
+
unit: 128
|
204 |
+
segment_size: 250
|
205 |
+
dropout: 0.1
|
206 |
+
mem_type: hc
|
207 |
+
seg_overlap: true
|
208 |
+
decoder: conv
|
209 |
+
decoder_conf:
|
210 |
+
channel: 64
|
211 |
+
kernel_size: 2
|
212 |
+
stride: 1
|
213 |
+
required:
|
214 |
+
- output_dir
|
215 |
+
version: 0.10.7a1
|
216 |
+
distributed: false
|
217 |
+
```
|
218 |
+
|
219 |
+
</details>
|
220 |
+
|
221 |
+
|
222 |
+
|
223 |
+
### Citing ESPnet
|
224 |
+
|
225 |
+
```BibTex
|
226 |
+
@inproceedings{watanabe2018espnet,
|
227 |
+
author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
|
228 |
+
title={{ESPnet}: End-to-End Speech Processing Toolkit},
|
229 |
+
year={2018},
|
230 |
+
booktitle={Proceedings of Interspeech},
|
231 |
+
pages={2207--2211},
|
232 |
+
doi={10.21437/Interspeech.2018-1456},
|
233 |
+
url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
|
234 |
+
}
|
235 |
+
|
236 |
+
|
237 |
+
@inproceedings{ESPnet-SE,
|
238 |
+
author = {Chenda Li and Jing Shi and Wangyou Zhang and Aswin Shanmugam Subramanian and Xuankai Chang and
|
239 |
+
Naoyuki Kamo and Moto Hira and Tomoki Hayashi and Christoph B{"{o}}ddeker and Zhuo Chen and Shinji Watanabe},
|
240 |
+
title = {ESPnet-SE: End-To-End Speech Enhancement and Separation Toolkit Designed for {ASR} Integration},
|
241 |
+
booktitle = {{IEEE} Spoken Language Technology Workshop, {SLT} 2021, Shenzhen, China, January 19-22, 2021},
|
242 |
+
pages = {785--792},
|
243 |
+
publisher = {{IEEE}},
|
244 |
+
year = {2021},
|
245 |
+
url = {https://doi.org/10.1109/SLT48900.2021.9383615},
|
246 |
+
doi = {10.1109/SLT48900.2021.9383615},
|
247 |
+
timestamp = {Mon, 12 Apr 2021 17:08:59 +0200},
|
248 |
+
biburl = {https://dblp.org/rec/conf/slt/Li0ZSCKHHBC021.bib},
|
249 |
+
bibsource = {dblp computer science bibliography, https://dblp.org}
|
250 |
+
}
|
251 |
+
|
252 |
+
|
253 |
+
```
|
254 |
+
|
255 |
+
or arXiv:
|
256 |
+
|
257 |
+
```bibtex
|
258 |
+
@misc{watanabe2018espnet,
|
259 |
+
title={ESPnet: End-to-End Speech Processing Toolkit},
|
260 |
+
author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
|
261 |
+
year={2018},
|
262 |
+
eprint={1804.00015},
|
263 |
+
archivePrefix={arXiv},
|
264 |
+
primaryClass={cs.CL}
|
265 |
+
}
|
266 |
+
```
|
exp/enh_stats_8k/train/feats_stats.npz
ADDED
Binary file (778 Bytes). View file
|
|
exp/enh_train_enh_skim_tasnet_noncausal_raw/146epoch.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:078d9c4cbc6a25a3d3df1e47a44355caa8be06d6fa48aed0e1c4884ac228b108
|
3 |
+
size 23716652
|
exp/enh_train_enh_skim_tasnet_noncausal_raw/RESULTS.md
ADDED
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
<!-- Generated by ./scripts/utils/show_enh_score.sh -->
|
2 |
+
# RESULTS
|
3 |
+
## Environments
|
4 |
+
- date: `Wed Feb 23 16:42:06 CST 2022`
|
5 |
+
- python version: `3.7.11 (default, Jul 27 2021, 14:32:16) [GCC 7.5.0]`
|
6 |
+
- espnet version: `espnet 0.10.7a1`
|
7 |
+
- pytorch version: `pytorch 1.8.1`
|
8 |
+
- Git hash: `ac3c10cfe4faf82c0bb30f8b32d9e8692363e0a9`
|
9 |
+
- Commit date: `Fri Feb 11 16:22:52 2022 +0800`
|
10 |
+
|
11 |
+
|
12 |
+
## ..
|
13 |
+
|
14 |
+
config: conf/tuning/train_enh_skim_tasnet_noncausal.yaml
|
15 |
+
|
16 |
+
|dataset|STOI|SAR|SDR|SIR|
|
17 |
+
|---|---|---|---|---|
|
18 |
+
|enhanced_cv_min_8k|0.96|19.17|18.70|29.56|
|
19 |
+
|enhanced_tt_min_8k|0.97|18.96|18.45|29.31|
|
20 |
+
|
exp/enh_train_enh_skim_tasnet_noncausal_raw/config.yaml
ADDED
@@ -0,0 +1,164 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
config: conf/tuning/train_enh_skim_tasnet_noncausal.yaml
|
2 |
+
print_config: false
|
3 |
+
log_level: INFO
|
4 |
+
dry_run: false
|
5 |
+
iterator_type: chunk
|
6 |
+
output_dir: exp/enh_train_enh_skim_tasnet_noncausal_raw
|
7 |
+
ngpu: 1
|
8 |
+
seed: 0
|
9 |
+
num_workers: 4
|
10 |
+
num_att_plot: 3
|
11 |
+
dist_backend: nccl
|
12 |
+
dist_init_method: env://
|
13 |
+
dist_world_size: null
|
14 |
+
dist_rank: null
|
15 |
+
local_rank: 0
|
16 |
+
dist_master_addr: null
|
17 |
+
dist_master_port: null
|
18 |
+
dist_launcher: null
|
19 |
+
multiprocessing_distributed: false
|
20 |
+
unused_parameters: false
|
21 |
+
sharded_ddp: false
|
22 |
+
cudnn_enabled: true
|
23 |
+
cudnn_benchmark: false
|
24 |
+
cudnn_deterministic: true
|
25 |
+
collect_stats: false
|
26 |
+
write_collected_feats: false
|
27 |
+
max_epoch: 150
|
28 |
+
patience: 20
|
29 |
+
val_scheduler_criterion:
|
30 |
+
- valid
|
31 |
+
- loss
|
32 |
+
early_stopping_criterion:
|
33 |
+
- valid
|
34 |
+
- loss
|
35 |
+
- min
|
36 |
+
best_model_criterion:
|
37 |
+
- - valid
|
38 |
+
- si_snr
|
39 |
+
- max
|
40 |
+
- - valid
|
41 |
+
- loss
|
42 |
+
- min
|
43 |
+
keep_nbest_models: 1
|
44 |
+
nbest_averaging_interval: 0
|
45 |
+
grad_clip: 5.0
|
46 |
+
grad_clip_type: 2.0
|
47 |
+
grad_noise: false
|
48 |
+
accum_grad: 1
|
49 |
+
no_forward_run: false
|
50 |
+
resume: true
|
51 |
+
train_dtype: float32
|
52 |
+
use_amp: false
|
53 |
+
log_interval: null
|
54 |
+
use_matplotlib: true
|
55 |
+
use_tensorboard: true
|
56 |
+
use_wandb: false
|
57 |
+
wandb_project: null
|
58 |
+
wandb_id: null
|
59 |
+
wandb_entity: null
|
60 |
+
wandb_name: null
|
61 |
+
wandb_model_log_interval: -1
|
62 |
+
detect_anomaly: false
|
63 |
+
pretrain_path: null
|
64 |
+
init_param: []
|
65 |
+
ignore_init_mismatch: false
|
66 |
+
freeze_param: []
|
67 |
+
num_iters_per_epoch: null
|
68 |
+
batch_size: 8
|
69 |
+
valid_batch_size: null
|
70 |
+
batch_bins: 1000000
|
71 |
+
valid_batch_bins: null
|
72 |
+
train_shape_file:
|
73 |
+
- exp/enh_stats_8k/train/speech_mix_shape
|
74 |
+
- exp/enh_stats_8k/train/speech_ref1_shape
|
75 |
+
- exp/enh_stats_8k/train/speech_ref2_shape
|
76 |
+
valid_shape_file:
|
77 |
+
- exp/enh_stats_8k/valid/speech_mix_shape
|
78 |
+
- exp/enh_stats_8k/valid/speech_ref1_shape
|
79 |
+
- exp/enh_stats_8k/valid/speech_ref2_shape
|
80 |
+
batch_type: folded
|
81 |
+
valid_batch_type: null
|
82 |
+
fold_length:
|
83 |
+
- 80000
|
84 |
+
- 80000
|
85 |
+
- 80000
|
86 |
+
sort_in_batch: descending
|
87 |
+
sort_batch: descending
|
88 |
+
multiple_iterator: false
|
89 |
+
chunk_length: 16000
|
90 |
+
chunk_shift_ratio: 0.5
|
91 |
+
num_cache_chunks: 1024
|
92 |
+
train_data_path_and_name_and_type:
|
93 |
+
- - dump/raw/tr_min_8k/wav.scp
|
94 |
+
- speech_mix
|
95 |
+
- sound
|
96 |
+
- - dump/raw/tr_min_8k/spk1.scp
|
97 |
+
- speech_ref1
|
98 |
+
- sound
|
99 |
+
- - dump/raw/tr_min_8k/spk2.scp
|
100 |
+
- speech_ref2
|
101 |
+
- sound
|
102 |
+
valid_data_path_and_name_and_type:
|
103 |
+
- - dump/raw/cv_min_8k/wav.scp
|
104 |
+
- speech_mix
|
105 |
+
- sound
|
106 |
+
- - dump/raw/cv_min_8k/spk1.scp
|
107 |
+
- speech_ref1
|
108 |
+
- sound
|
109 |
+
- - dump/raw/cv_min_8k/spk2.scp
|
110 |
+
- speech_ref2
|
111 |
+
- sound
|
112 |
+
allow_variable_data_keys: false
|
113 |
+
max_cache_size: 0.0
|
114 |
+
max_cache_fd: 32
|
115 |
+
valid_max_cache_size: null
|
116 |
+
optim: adam
|
117 |
+
optim_conf:
|
118 |
+
lr: 0.001
|
119 |
+
eps: 1.0e-08
|
120 |
+
weight_decay: 0
|
121 |
+
scheduler: reducelronplateau
|
122 |
+
scheduler_conf:
|
123 |
+
mode: min
|
124 |
+
factor: 0.7
|
125 |
+
patience: 1
|
126 |
+
init: xavier_uniform
|
127 |
+
model_conf:
|
128 |
+
stft_consistency: false
|
129 |
+
loss_type: mask_mse
|
130 |
+
mask_type: null
|
131 |
+
criterions:
|
132 |
+
- name: si_snr
|
133 |
+
conf:
|
134 |
+
eps: 1.0e-07
|
135 |
+
wrapper: pit
|
136 |
+
wrapper_conf:
|
137 |
+
weight: 1.0
|
138 |
+
independent_perm: true
|
139 |
+
use_preprocessor: false
|
140 |
+
encoder: conv
|
141 |
+
encoder_conf:
|
142 |
+
channel: 64
|
143 |
+
kernel_size: 2
|
144 |
+
stride: 1
|
145 |
+
separator: skim
|
146 |
+
separator_conf:
|
147 |
+
causal: false
|
148 |
+
num_spk: 2
|
149 |
+
layer: 6
|
150 |
+
nonlinear: relu
|
151 |
+
unit: 128
|
152 |
+
segment_size: 250
|
153 |
+
dropout: 0.1
|
154 |
+
mem_type: hc
|
155 |
+
seg_overlap: true
|
156 |
+
decoder: conv
|
157 |
+
decoder_conf:
|
158 |
+
channel: 64
|
159 |
+
kernel_size: 2
|
160 |
+
stride: 1
|
161 |
+
required:
|
162 |
+
- output_dir
|
163 |
+
version: 0.10.7a1
|
164 |
+
distributed: false
|
exp/enh_train_enh_skim_tasnet_noncausal_raw/images/backward_time.png
ADDED
exp/enh_train_enh_skim_tasnet_noncausal_raw/images/forward_time.png
ADDED
exp/enh_train_enh_skim_tasnet_noncausal_raw/images/gpu_max_cached_mem_GB.png
ADDED
exp/enh_train_enh_skim_tasnet_noncausal_raw/images/iter_time.png
ADDED
exp/enh_train_enh_skim_tasnet_noncausal_raw/images/loss.png
ADDED
exp/enh_train_enh_skim_tasnet_noncausal_raw/images/optim0_lr0.png
ADDED
exp/enh_train_enh_skim_tasnet_noncausal_raw/images/optim_step_time.png
ADDED
exp/enh_train_enh_skim_tasnet_noncausal_raw/images/si_snr_loss.png
ADDED
exp/enh_train_enh_skim_tasnet_noncausal_raw/images/train_time.png
ADDED
meta.yaml
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
espnet: 0.10.7a1
|
2 |
+
files:
|
3 |
+
model_file: exp/enh_train_enh_skim_tasnet_noncausal_raw/146epoch.pth
|
4 |
+
python: "3.7.11 (default, Jul 27 2021, 14:32:16) \n[GCC 7.5.0]"
|
5 |
+
timestamp: 1645606563.393269
|
6 |
+
torch: 1.8.1
|
7 |
+
yaml_files:
|
8 |
+
train_config: exp/enh_train_enh_skim_tasnet_noncausal_raw/config.yaml
|