Update model
Browse files- README.md +476 -0
- data/eu_token_list/bpe_unigram150/bpe.model +3 -0
- exp/asr_stats_raw_eu_bpe150_sp/train/feats_stats.npz +3 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/RESULTS.md +48 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/config.yaml +358 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/backward_time.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/cer_ctc.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/cer_transducer.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/clip.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/forward_time.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/gpu_max_cached_mem_GB.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/grad_norm.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/iter_time.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/loss.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/loss_ctc.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/loss_scale.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/loss_transducer.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/optim0_lr0.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/optim_step_time.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/train_time.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/wer_transducer.png +0 -0
- exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/valid.loss.ave_10best.pth +3 -0
- exp/lm_train_lm_eu_bpe150/20epoch.pth +3 -0
- exp/lm_train_lm_eu_bpe150/config.yaml +274 -0
- exp/lm_train_lm_eu_bpe150/images/backward_time.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/clip.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/forward_time.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/gpu_max_cached_mem_GB.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/grad_norm.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/iter_time.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/loss.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/loss_scale.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/optim0_lr0.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/optim_step_time.png +0 -0
- exp/lm_train_lm_eu_bpe150/images/train_time.png +0 -0
- exp/lm_train_lm_eu_bpe150/perplexity_test/ppl +1 -0
- meta.yaml +11 -0
README.md
CHANGED
@@ -1,3 +1,479 @@
|
|
1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: cc-by-4.0
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
tags:
|
3 |
+
- espnet
|
4 |
+
- audio
|
5 |
+
- automatic-speech-recognition
|
6 |
+
language: eu
|
7 |
+
datasets:
|
8 |
+
- commonvoice
|
9 |
license: cc-by-4.0
|
10 |
---
|
11 |
+
|
12 |
+
## ESPnet2 ASR model
|
13 |
+
|
14 |
+
### `espnet/zuazo_commonvoice_asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp`
|
15 |
+
|
16 |
+
This model was trained by Xabier de Zuazo using commonvoice recipe in [espnet](https://github.com/espnet/espnet/).
|
17 |
+
|
18 |
+
### Demo: How to use in ESPnet2
|
19 |
+
|
20 |
+
Follow the [ESPnet installation instructions](https://espnet.github.io/espnet/installation.html)
|
21 |
+
if you haven't done that already.
|
22 |
+
|
23 |
+
```bash
|
24 |
+
cd espnet
|
25 |
+
git checkout 5d0758e2a7063b82d1f10a8ac2de98eb6cf8a352
|
26 |
+
pip install -e .
|
27 |
+
cd egs2/commonvoice/asr1.eu.transducer_conformer5
|
28 |
+
./run.sh --skip_data_prep false --skip_train true --download_model espnet/zuazo_commonvoice_asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp
|
29 |
+
```
|
30 |
+
|
31 |
+
<!-- Generated by scripts/utils/show_asr_result.sh -->
|
32 |
+
# RESULTS
|
33 |
+
## Environments
|
34 |
+
- date: `Wed Oct 25 12:18:16 CEST 2023`
|
35 |
+
- python version: `3.8.17 | packaged by conda-forge | (default, Jun 16 2023, 07:06:00) [GCC 11.4.0]`
|
36 |
+
- espnet version: `espnet 202308`
|
37 |
+
- pytorch version: `pytorch 2.0.1`
|
38 |
+
- Git hash: `5d0758e2a7063b82d1f10a8ac2de98eb6cf8a352`
|
39 |
+
- Commit date: `Wed Aug 30 18:03:42 2023 -0400`
|
40 |
+
|
41 |
+
## exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp
|
42 |
+
### WER
|
43 |
+
|
44 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
45 |
+
|---|---|---|---|---|---|---|---|---|
|
46 |
+
|decode_transducer_lm_lm_train_lm_eu_bpe150_valid.loss.ave_asr_model_valid.loss.ave/test_eu|6640|49267|92.9|6.6|0.5|0.8|8.0|33.6|
|
47 |
+
|
48 |
+
### CER
|
49 |
+
|
50 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
51 |
+
|---|---|---|---|---|---|---|---|---|
|
52 |
+
|decode_transducer_lm_lm_train_lm_eu_bpe150_valid.loss.ave_asr_model_valid.loss.ave/test_eu|6640|373913|98.7|0.6|0.8|0.4|1.7|33.6|
|
53 |
+
|
54 |
+
### TER
|
55 |
+
|
56 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
57 |
+
|---|---|---|---|---|---|---|---|---|
|
58 |
+
|decode_transducer_lm_lm_train_lm_eu_bpe150_valid.loss.ave_asr_model_valid.loss.ave/test_eu|6640|208360|97.3|1.5|1.2|0.5|3.2|33.6|
|
59 |
+
|
60 |
+
## exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/decode_transducer_lm_lm_train_lm_eu_bpe150_valid.loss.ave_asr_model_valid.loss.ave
|
61 |
+
### WER
|
62 |
+
|
63 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
64 |
+
|---|---|---|---|---|---|---|---|---|
|
65 |
+
|org/dev_eu|6640|49505|93.5|6.1|0.4|0.8|7.3|31.7|
|
66 |
+
|
67 |
+
### CER
|
68 |
+
|
69 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
70 |
+
|---|---|---|---|---|---|---|---|---|
|
71 |
+
|org/dev_eu|6640|376502|98.9|0.5|0.6|0.3|1.4|31.7|
|
72 |
+
|
73 |
+
### TER
|
74 |
+
|
75 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
76 |
+
|---|---|---|---|---|---|---|---|---|
|
77 |
+
|org/dev_eu|6640|209465|97.7|1.3|1.1|0.4|2.8|31.7|
|
78 |
+
|
79 |
+
## ASR config
|
80 |
+
|
81 |
+
<details><summary>expand</summary>
|
82 |
+
|
83 |
+
```
|
84 |
+
config: conf/tuning/train_asr_transducer_conformer5.yaml
|
85 |
+
print_config: false
|
86 |
+
log_level: INFO
|
87 |
+
drop_last_iter: false
|
88 |
+
dry_run: false
|
89 |
+
iterator_type: sequence
|
90 |
+
valid_iterator_type: null
|
91 |
+
output_dir: exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp
|
92 |
+
ngpu: 1
|
93 |
+
seed: 2022
|
94 |
+
num_workers: 8
|
95 |
+
num_att_plot: 3
|
96 |
+
dist_backend: nccl
|
97 |
+
dist_init_method: env://
|
98 |
+
dist_world_size: null
|
99 |
+
dist_rank: null
|
100 |
+
local_rank: 0
|
101 |
+
dist_master_addr: null
|
102 |
+
dist_master_port: null
|
103 |
+
dist_launcher: null
|
104 |
+
multiprocessing_distributed: false
|
105 |
+
unused_parameters: false
|
106 |
+
sharded_ddp: false
|
107 |
+
cudnn_enabled: true
|
108 |
+
cudnn_benchmark: false
|
109 |
+
cudnn_deterministic: true
|
110 |
+
collect_stats: false
|
111 |
+
write_collected_feats: false
|
112 |
+
max_epoch: 50
|
113 |
+
patience: null
|
114 |
+
val_scheduler_criterion:
|
115 |
+
- valid
|
116 |
+
- loss
|
117 |
+
early_stopping_criterion:
|
118 |
+
- valid
|
119 |
+
- loss
|
120 |
+
- min
|
121 |
+
best_model_criterion:
|
122 |
+
- - valid
|
123 |
+
- loss
|
124 |
+
- min
|
125 |
+
keep_nbest_models: 10
|
126 |
+
nbest_averaging_interval: 0
|
127 |
+
grad_clip: 3
|
128 |
+
grad_clip_type: 2.0
|
129 |
+
grad_noise: false
|
130 |
+
accum_grad: 1
|
131 |
+
no_forward_run: false
|
132 |
+
resume: true
|
133 |
+
train_dtype: float32
|
134 |
+
use_amp: false
|
135 |
+
log_interval: null
|
136 |
+
use_matplotlib: true
|
137 |
+
use_tensorboard: true
|
138 |
+
create_graph_in_tensorboard: false
|
139 |
+
use_wandb: false
|
140 |
+
wandb_project: null
|
141 |
+
wandb_id: null
|
142 |
+
wandb_entity: null
|
143 |
+
wandb_name: null
|
144 |
+
wandb_model_log_interval: -1
|
145 |
+
detect_anomaly: false
|
146 |
+
pretrain_path: null
|
147 |
+
init_param: []
|
148 |
+
ignore_init_mismatch: false
|
149 |
+
freeze_param: []
|
150 |
+
num_iters_per_epoch: null
|
151 |
+
batch_size: 20
|
152 |
+
valid_batch_size: null
|
153 |
+
batch_bins: 10000000
|
154 |
+
valid_batch_bins: null
|
155 |
+
train_shape_file:
|
156 |
+
- exp/asr_stats_raw_eu_bpe150_sp/train/speech_shape
|
157 |
+
- exp/asr_stats_raw_eu_bpe150_sp/train/text_shape.bpe
|
158 |
+
valid_shape_file:
|
159 |
+
- exp/asr_stats_raw_eu_bpe150_sp/valid/speech_shape
|
160 |
+
- exp/asr_stats_raw_eu_bpe150_sp/valid/text_shape.bpe
|
161 |
+
batch_type: numel
|
162 |
+
valid_batch_type: null
|
163 |
+
fold_length:
|
164 |
+
- 80000
|
165 |
+
- 150
|
166 |
+
sort_in_batch: descending
|
167 |
+
shuffle_within_batch: false
|
168 |
+
sort_batch: descending
|
169 |
+
multiple_iterator: false
|
170 |
+
chunk_length: 500
|
171 |
+
chunk_shift_ratio: 0.5
|
172 |
+
num_cache_chunks: 1024
|
173 |
+
chunk_excluded_key_prefixes: []
|
174 |
+
train_data_path_and_name_and_type:
|
175 |
+
- - dump/raw/train_eu_sp/wav.scp
|
176 |
+
- speech
|
177 |
+
- sound
|
178 |
+
- - dump/raw/train_eu_sp/text
|
179 |
+
- text
|
180 |
+
- text
|
181 |
+
valid_data_path_and_name_and_type:
|
182 |
+
- - dump/raw/dev_eu/wav.scp
|
183 |
+
- speech
|
184 |
+
- sound
|
185 |
+
- - dump/raw/dev_eu/text
|
186 |
+
- text
|
187 |
+
- text
|
188 |
+
allow_variable_data_keys: false
|
189 |
+
max_cache_size: 0.0
|
190 |
+
max_cache_fd: 32
|
191 |
+
valid_max_cache_size: null
|
192 |
+
exclude_weight_decay: false
|
193 |
+
exclude_weight_decay_conf: {}
|
194 |
+
optim: adam
|
195 |
+
optim_conf:
|
196 |
+
lr: 4.0
|
197 |
+
scheduler: noamlr
|
198 |
+
scheduler_conf:
|
199 |
+
model_size: 256
|
200 |
+
warmup_steps: 25000
|
201 |
+
token_list:
|
202 |
+
- <blank>
|
203 |
+
- <unk>
|
204 |
+
- A
|
205 |
+
- ▁
|
206 |
+
- I
|
207 |
+
- E
|
208 |
+
- Z
|
209 |
+
- .
|
210 |
+
- R
|
211 |
+
- N
|
212 |
+
- U
|
213 |
+
- S
|
214 |
+
- O
|
215 |
+
- T
|
216 |
+
- KO
|
217 |
+
- K
|
218 |
+
- ▁E
|
219 |
+
- TU
|
220 |
+
- TE
|
221 |
+
- RA
|
222 |
+
- EN
|
223 |
+
- L
|
224 |
+
- ','
|
225 |
+
- LA
|
226 |
+
- TA
|
227 |
+
- AK
|
228 |
+
- ▁A
|
229 |
+
- AN
|
230 |
+
- ▁DA
|
231 |
+
- RE
|
232 |
+
- KA
|
233 |
+
- P
|
234 |
+
- GO
|
235 |
+
- IN
|
236 |
+
- B
|
237 |
+
- M
|
238 |
+
- ▁DU
|
239 |
+
- RI
|
240 |
+
- GU
|
241 |
+
- ▁ETA
|
242 |
+
- D
|
243 |
+
- ER
|
244 |
+
- UR
|
245 |
+
- ▁BA
|
246 |
+
- ▁P
|
247 |
+
- H
|
248 |
+
- MA
|
249 |
+
- ▁G
|
250 |
+
- ▁I
|
251 |
+
- ▁HA
|
252 |
+
- TZEN
|
253 |
+
- LE
|
254 |
+
- ▁EZ
|
255 |
+
- ▁O
|
256 |
+
- EK
|
257 |
+
- GI
|
258 |
+
- ▁BAT
|
259 |
+
- DA
|
260 |
+
- DU
|
261 |
+
- TZA
|
262 |
+
- KI
|
263 |
+
- DI
|
264 |
+
- RO
|
265 |
+
- ▁GA
|
266 |
+
- REN
|
267 |
+
- AR
|
268 |
+
- TEN
|
269 |
+
- GA
|
270 |
+
- TIK
|
271 |
+
- RRI
|
272 |
+
- ▁BI
|
273 |
+
- LI
|
274 |
+
- ▁BER
|
275 |
+
- G
|
276 |
+
- ▁AR
|
277 |
+
- TO
|
278 |
+
- ERA
|
279 |
+
- AREN
|
280 |
+
- ▁ZI
|
281 |
+
- ▁DE
|
282 |
+
- ▁BE
|
283 |
+
- X
|
284 |
+
- BA
|
285 |
+
- ▁DI
|
286 |
+
- ▁IZAN
|
287 |
+
- ▁ZE
|
288 |
+
- ETAN
|
289 |
+
- ▁ZEN
|
290 |
+
- EAN
|
291 |
+
- IA
|
292 |
+
- ▁JA
|
293 |
+
- ▁ERE
|
294 |
+
- ▁DITU
|
295 |
+
- ▁ZA
|
296 |
+
- ▁ERA
|
297 |
+
- LO
|
298 |
+
- ▁HOR
|
299 |
+
- NTZ
|
300 |
+
- ▁DIRA
|
301 |
+
- MEN
|
302 |
+
- ▁HI
|
303 |
+
- ▁F
|
304 |
+
- F
|
305 |
+
- LDE
|
306 |
+
- ZIO
|
307 |
+
- '?'
|
308 |
+
- ▁ZU
|
309 |
+
- '-'
|
310 |
+
- DO
|
311 |
+
- ▁EGIN
|
312 |
+
- TZEKO
|
313 |
+
- ▁BEHAR
|
314 |
+
- TZI
|
315 |
+
- BIL
|
316 |
+
- ▁IN
|
317 |
+
- RIK
|
318 |
+
- ▁HORI
|
319 |
+
- ▁SA
|
320 |
+
- ▁NA
|
321 |
+
- BIDE
|
322 |
+
- ▁KON
|
323 |
+
- ▁HE
|
324 |
+
- ▁ZUEN
|
325 |
+
- ▁MU
|
326 |
+
- ▁BESTE
|
327 |
+
- ▁SO
|
328 |
+
- ▁HERRI
|
329 |
+
- ▁IKAS
|
330 |
+
- ▁NO
|
331 |
+
- ▁ALD
|
332 |
+
- ▁NI
|
333 |
+
- ▁TX
|
334 |
+
- ABE
|
335 |
+
- KETA
|
336 |
+
- ▁BAINA
|
337 |
+
- C
|
338 |
+
- '!'
|
339 |
+
- V
|
340 |
+
- Y
|
341 |
+
- ':'
|
342 |
+
- ;
|
343 |
+
- '"'
|
344 |
+
- Í
|
345 |
+
- Q
|
346 |
+
- Ñ
|
347 |
+
- W
|
348 |
+
- J
|
349 |
+
- ‘
|
350 |
+
- ’
|
351 |
+
- <sos/eos>
|
352 |
+
init: xavier_uniform
|
353 |
+
input_size: null
|
354 |
+
ctc_conf:
|
355 |
+
dropout_rate: 0.0
|
356 |
+
ctc_type: builtin
|
357 |
+
reduce: true
|
358 |
+
ignore_nan_grad: null
|
359 |
+
zero_infinity: true
|
360 |
+
joint_net_conf:
|
361 |
+
joint_space_size: 640
|
362 |
+
use_preprocessor: true
|
363 |
+
token_type: bpe
|
364 |
+
bpemodel: data/eu_token_list/bpe_unigram150/bpe.model
|
365 |
+
non_linguistic_symbols: null
|
366 |
+
cleaner: null
|
367 |
+
g2p: null
|
368 |
+
speech_volume_normalize: null
|
369 |
+
rir_scp: null
|
370 |
+
rir_apply_prob: 1.0
|
371 |
+
noise_scp: null
|
372 |
+
noise_apply_prob: 1.0
|
373 |
+
noise_db_range: '13_15'
|
374 |
+
short_noise_thres: 0.5
|
375 |
+
aux_ctc_tasks: []
|
376 |
+
frontend: default
|
377 |
+
frontend_conf:
|
378 |
+
n_fft: 512
|
379 |
+
win_length: 400
|
380 |
+
hop_length: 160
|
381 |
+
fs: 16k
|
382 |
+
specaug: specaug
|
383 |
+
specaug_conf:
|
384 |
+
apply_time_warp: true
|
385 |
+
time_warp_window: 5
|
386 |
+
time_warp_mode: bicubic
|
387 |
+
apply_freq_mask: true
|
388 |
+
freq_mask_width_range:
|
389 |
+
- 0
|
390 |
+
- 30
|
391 |
+
num_freq_mask: 2
|
392 |
+
apply_time_mask: true
|
393 |
+
time_mask_width_range:
|
394 |
+
- 0
|
395 |
+
- 40
|
396 |
+
num_time_mask: 2
|
397 |
+
normalize: global_mvn
|
398 |
+
normalize_conf:
|
399 |
+
stats_file: exp/asr_stats_raw_eu_bpe150_sp/train/feats_stats.npz
|
400 |
+
model: espnet
|
401 |
+
model_conf:
|
402 |
+
ctc_weight: 0.3
|
403 |
+
lsm_weight: 0.1
|
404 |
+
length_normalized_loss: false
|
405 |
+
report_cer: true
|
406 |
+
report_wer: true
|
407 |
+
preencoder: null
|
408 |
+
preencoder_conf: {}
|
409 |
+
encoder: conformer
|
410 |
+
encoder_conf:
|
411 |
+
input_layer: conv2d
|
412 |
+
num_blocks: 12
|
413 |
+
linear_units: 2048
|
414 |
+
dropout_rate: 0.1
|
415 |
+
output_size: 256
|
416 |
+
attention_heads: 4
|
417 |
+
attention_dropout_rate: 0.0
|
418 |
+
positional_dropout_rate: 0.1
|
419 |
+
pos_enc_layer_type: rel_pos
|
420 |
+
selfattention_layer_type: rel_selfattn
|
421 |
+
activation_type: swish
|
422 |
+
rel_pos_type: latest
|
423 |
+
macaron_style: true
|
424 |
+
use_cnn_module: true
|
425 |
+
cnn_module_kernel: 15
|
426 |
+
postencoder: null
|
427 |
+
postencoder_conf: {}
|
428 |
+
decoder: transducer
|
429 |
+
decoder_conf:
|
430 |
+
rnn_type: lstm
|
431 |
+
num_layers: 1
|
432 |
+
hidden_size: 640
|
433 |
+
dropout: 0.1
|
434 |
+
dropout_embed: 0.2
|
435 |
+
preprocessor: default
|
436 |
+
preprocessor_conf: {}
|
437 |
+
required:
|
438 |
+
- output_dir
|
439 |
+
- token_list
|
440 |
+
version: '202308'
|
441 |
+
distributed: false
|
442 |
+
```
|
443 |
+
|
444 |
+
</details>
|
445 |
+
|
446 |
+
|
447 |
+
|
448 |
+
### Citing ESPnet
|
449 |
+
|
450 |
+
```BibTex
|
451 |
+
@inproceedings{watanabe2018espnet,
|
452 |
+
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},
|
453 |
+
title={{ESPnet}: End-to-End Speech Processing Toolkit},
|
454 |
+
year={2018},
|
455 |
+
booktitle={Proceedings of Interspeech},
|
456 |
+
pages={2207--2211},
|
457 |
+
doi={10.21437/Interspeech.2018-1456},
|
458 |
+
url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
|
459 |
+
}
|
460 |
+
|
461 |
+
|
462 |
+
|
463 |
+
|
464 |
+
|
465 |
+
|
466 |
+
```
|
467 |
+
|
468 |
+
or arXiv:
|
469 |
+
|
470 |
+
```bibtex
|
471 |
+
@misc{watanabe2018espnet,
|
472 |
+
title={ESPnet: End-to-End Speech Processing Toolkit},
|
473 |
+
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},
|
474 |
+
year={2018},
|
475 |
+
eprint={1804.00015},
|
476 |
+
archivePrefix={arXiv},
|
477 |
+
primaryClass={cs.CL}
|
478 |
+
}
|
479 |
+
```
|
data/eu_token_list/bpe_unigram150/bpe.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:26fa5a483ff569257301f74a2df5994caac6911d403a64afaf3e8678c02e7f64
|
3 |
+
size 239526
|
exp/asr_stats_raw_eu_bpe150_sp/train/feats_stats.npz
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:047a8feae50058a2b37d1f99320350993783f139d659f4f1b43b31a14769b42c
|
3 |
+
size 1402
|
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/RESULTS.md
ADDED
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
<!-- Generated by scripts/utils/show_asr_result.sh -->
|
2 |
+
# RESULTS
|
3 |
+
## Environments
|
4 |
+
- date: `Wed Oct 25 12:18:16 CEST 2023`
|
5 |
+
- python version: `3.8.17 | packaged by conda-forge | (default, Jun 16 2023, 07:06:00) [GCC 11.4.0]`
|
6 |
+
- espnet version: `espnet 202308`
|
7 |
+
- pytorch version: `pytorch 2.0.1`
|
8 |
+
- Git hash: `5d0758e2a7063b82d1f10a8ac2de98eb6cf8a352`
|
9 |
+
- Commit date: `Wed Aug 30 18:03:42 2023 -0400`
|
10 |
+
|
11 |
+
## exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp
|
12 |
+
### WER
|
13 |
+
|
14 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
15 |
+
|---|---|---|---|---|---|---|---|---|
|
16 |
+
|decode_transducer_lm_lm_train_lm_eu_bpe150_valid.loss.ave_asr_model_valid.loss.ave/test_eu|6640|49267|92.9|6.6|0.5|0.8|8.0|33.6|
|
17 |
+
|
18 |
+
### CER
|
19 |
+
|
20 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
21 |
+
|---|---|---|---|---|---|---|---|---|
|
22 |
+
|decode_transducer_lm_lm_train_lm_eu_bpe150_valid.loss.ave_asr_model_valid.loss.ave/test_eu|6640|373913|98.7|0.6|0.8|0.4|1.7|33.6|
|
23 |
+
|
24 |
+
### TER
|
25 |
+
|
26 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
27 |
+
|---|---|---|---|---|---|---|---|---|
|
28 |
+
|decode_transducer_lm_lm_train_lm_eu_bpe150_valid.loss.ave_asr_model_valid.loss.ave/test_eu|6640|208360|97.3|1.5|1.2|0.5|3.2|33.6|
|
29 |
+
|
30 |
+
## exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/decode_transducer_lm_lm_train_lm_eu_bpe150_valid.loss.ave_asr_model_valid.loss.ave
|
31 |
+
### WER
|
32 |
+
|
33 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
34 |
+
|---|---|---|---|---|---|---|---|---|
|
35 |
+
|org/dev_eu|6640|49505|93.5|6.1|0.4|0.8|7.3|31.7|
|
36 |
+
|
37 |
+
### CER
|
38 |
+
|
39 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
40 |
+
|---|---|---|---|---|---|---|---|---|
|
41 |
+
|org/dev_eu|6640|376502|98.9|0.5|0.6|0.3|1.4|31.7|
|
42 |
+
|
43 |
+
### TER
|
44 |
+
|
45 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
46 |
+
|---|---|---|---|---|---|---|---|---|
|
47 |
+
|org/dev_eu|6640|209465|97.7|1.3|1.1|0.4|2.8|31.7|
|
48 |
+
|
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/config.yaml
ADDED
@@ -0,0 +1,358 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
config: conf/tuning/train_asr_transducer_conformer5.yaml
|
2 |
+
print_config: false
|
3 |
+
log_level: INFO
|
4 |
+
drop_last_iter: false
|
5 |
+
dry_run: false
|
6 |
+
iterator_type: sequence
|
7 |
+
valid_iterator_type: null
|
8 |
+
output_dir: exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp
|
9 |
+
ngpu: 1
|
10 |
+
seed: 2022
|
11 |
+
num_workers: 8
|
12 |
+
num_att_plot: 3
|
13 |
+
dist_backend: nccl
|
14 |
+
dist_init_method: env://
|
15 |
+
dist_world_size: null
|
16 |
+
dist_rank: null
|
17 |
+
local_rank: 0
|
18 |
+
dist_master_addr: null
|
19 |
+
dist_master_port: null
|
20 |
+
dist_launcher: null
|
21 |
+
multiprocessing_distributed: false
|
22 |
+
unused_parameters: false
|
23 |
+
sharded_ddp: false
|
24 |
+
cudnn_enabled: true
|
25 |
+
cudnn_benchmark: false
|
26 |
+
cudnn_deterministic: true
|
27 |
+
collect_stats: false
|
28 |
+
write_collected_feats: false
|
29 |
+
max_epoch: 50
|
30 |
+
patience: null
|
31 |
+
val_scheduler_criterion:
|
32 |
+
- valid
|
33 |
+
- loss
|
34 |
+
early_stopping_criterion:
|
35 |
+
- valid
|
36 |
+
- loss
|
37 |
+
- min
|
38 |
+
best_model_criterion:
|
39 |
+
- - valid
|
40 |
+
- loss
|
41 |
+
- min
|
42 |
+
keep_nbest_models: 10
|
43 |
+
nbest_averaging_interval: 0
|
44 |
+
grad_clip: 3
|
45 |
+
grad_clip_type: 2.0
|
46 |
+
grad_noise: false
|
47 |
+
accum_grad: 1
|
48 |
+
no_forward_run: false
|
49 |
+
resume: true
|
50 |
+
train_dtype: float32
|
51 |
+
use_amp: false
|
52 |
+
log_interval: null
|
53 |
+
use_matplotlib: true
|
54 |
+
use_tensorboard: true
|
55 |
+
create_graph_in_tensorboard: false
|
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: 20
|
69 |
+
valid_batch_size: null
|
70 |
+
batch_bins: 10000000
|
71 |
+
valid_batch_bins: null
|
72 |
+
train_shape_file:
|
73 |
+
- exp/asr_stats_raw_eu_bpe150_sp/train/speech_shape
|
74 |
+
- exp/asr_stats_raw_eu_bpe150_sp/train/text_shape.bpe
|
75 |
+
valid_shape_file:
|
76 |
+
- exp/asr_stats_raw_eu_bpe150_sp/valid/speech_shape
|
77 |
+
- exp/asr_stats_raw_eu_bpe150_sp/valid/text_shape.bpe
|
78 |
+
batch_type: numel
|
79 |
+
valid_batch_type: null
|
80 |
+
fold_length:
|
81 |
+
- 80000
|
82 |
+
- 150
|
83 |
+
sort_in_batch: descending
|
84 |
+
shuffle_within_batch: false
|
85 |
+
sort_batch: descending
|
86 |
+
multiple_iterator: false
|
87 |
+
chunk_length: 500
|
88 |
+
chunk_shift_ratio: 0.5
|
89 |
+
num_cache_chunks: 1024
|
90 |
+
chunk_excluded_key_prefixes: []
|
91 |
+
train_data_path_and_name_and_type:
|
92 |
+
- - dump/raw/train_eu_sp/wav.scp
|
93 |
+
- speech
|
94 |
+
- sound
|
95 |
+
- - dump/raw/train_eu_sp/text
|
96 |
+
- text
|
97 |
+
- text
|
98 |
+
valid_data_path_and_name_and_type:
|
99 |
+
- - dump/raw/dev_eu/wav.scp
|
100 |
+
- speech
|
101 |
+
- sound
|
102 |
+
- - dump/raw/dev_eu/text
|
103 |
+
- text
|
104 |
+
- text
|
105 |
+
allow_variable_data_keys: false
|
106 |
+
max_cache_size: 0.0
|
107 |
+
max_cache_fd: 32
|
108 |
+
valid_max_cache_size: null
|
109 |
+
exclude_weight_decay: false
|
110 |
+
exclude_weight_decay_conf: {}
|
111 |
+
optim: adam
|
112 |
+
optim_conf:
|
113 |
+
lr: 4.0
|
114 |
+
scheduler: noamlr
|
115 |
+
scheduler_conf:
|
116 |
+
model_size: 256
|
117 |
+
warmup_steps: 25000
|
118 |
+
token_list:
|
119 |
+
- <blank>
|
120 |
+
- <unk>
|
121 |
+
- A
|
122 |
+
- ▁
|
123 |
+
- I
|
124 |
+
- E
|
125 |
+
- Z
|
126 |
+
- .
|
127 |
+
- R
|
128 |
+
- N
|
129 |
+
- U
|
130 |
+
- S
|
131 |
+
- O
|
132 |
+
- T
|
133 |
+
- KO
|
134 |
+
- K
|
135 |
+
- ▁E
|
136 |
+
- TU
|
137 |
+
- TE
|
138 |
+
- RA
|
139 |
+
- EN
|
140 |
+
- L
|
141 |
+
- ','
|
142 |
+
- LA
|
143 |
+
- TA
|
144 |
+
- AK
|
145 |
+
- ▁A
|
146 |
+
- AN
|
147 |
+
- ▁DA
|
148 |
+
- RE
|
149 |
+
- KA
|
150 |
+
- P
|
151 |
+
- GO
|
152 |
+
- IN
|
153 |
+
- B
|
154 |
+
- M
|
155 |
+
- ▁DU
|
156 |
+
- RI
|
157 |
+
- GU
|
158 |
+
- ▁ETA
|
159 |
+
- D
|
160 |
+
- ER
|
161 |
+
- UR
|
162 |
+
- ▁BA
|
163 |
+
- ▁P
|
164 |
+
- H
|
165 |
+
- MA
|
166 |
+
- ▁G
|
167 |
+
- ▁I
|
168 |
+
- ▁HA
|
169 |
+
- TZEN
|
170 |
+
- LE
|
171 |
+
- ▁EZ
|
172 |
+
- ▁O
|
173 |
+
- EK
|
174 |
+
- GI
|
175 |
+
- ▁BAT
|
176 |
+
- DA
|
177 |
+
- DU
|
178 |
+
- TZA
|
179 |
+
- KI
|
180 |
+
- DI
|
181 |
+
- RO
|
182 |
+
- ▁GA
|
183 |
+
- REN
|
184 |
+
- AR
|
185 |
+
- TEN
|
186 |
+
- GA
|
187 |
+
- TIK
|
188 |
+
- RRI
|
189 |
+
- ▁BI
|
190 |
+
- LI
|
191 |
+
- ▁BER
|
192 |
+
- G
|
193 |
+
- ▁AR
|
194 |
+
- TO
|
195 |
+
- ERA
|
196 |
+
- AREN
|
197 |
+
- ▁ZI
|
198 |
+
- ▁DE
|
199 |
+
- ▁BE
|
200 |
+
- X
|
201 |
+
- BA
|
202 |
+
- ▁DI
|
203 |
+
- ▁IZAN
|
204 |
+
- ▁ZE
|
205 |
+
- ETAN
|
206 |
+
- ▁ZEN
|
207 |
+
- EAN
|
208 |
+
- IA
|
209 |
+
- ▁JA
|
210 |
+
- ▁ERE
|
211 |
+
- ▁DITU
|
212 |
+
- ▁ZA
|
213 |
+
- ▁ERA
|
214 |
+
- LO
|
215 |
+
- ▁HOR
|
216 |
+
- NTZ
|
217 |
+
- ▁DIRA
|
218 |
+
- MEN
|
219 |
+
- ▁HI
|
220 |
+
- ▁F
|
221 |
+
- F
|
222 |
+
- LDE
|
223 |
+
- ZIO
|
224 |
+
- '?'
|
225 |
+
- ▁ZU
|
226 |
+
- '-'
|
227 |
+
- DO
|
228 |
+
- ▁EGIN
|
229 |
+
- TZEKO
|
230 |
+
- ▁BEHAR
|
231 |
+
- TZI
|
232 |
+
- BIL
|
233 |
+
- ▁IN
|
234 |
+
- RIK
|
235 |
+
- ▁HORI
|
236 |
+
- ▁SA
|
237 |
+
- ▁NA
|
238 |
+
- BIDE
|
239 |
+
- ▁KON
|
240 |
+
- ▁HE
|
241 |
+
- ▁ZUEN
|
242 |
+
- ▁MU
|
243 |
+
- ▁BESTE
|
244 |
+
- ▁SO
|
245 |
+
- ▁HERRI
|
246 |
+
- ▁IKAS
|
247 |
+
- ▁NO
|
248 |
+
- ▁ALD
|
249 |
+
- ▁NI
|
250 |
+
- ▁TX
|
251 |
+
- ABE
|
252 |
+
- KETA
|
253 |
+
- ▁BAINA
|
254 |
+
- C
|
255 |
+
- '!'
|
256 |
+
- V
|
257 |
+
- Y
|
258 |
+
- ':'
|
259 |
+
- ;
|
260 |
+
- '"'
|
261 |
+
- Í
|
262 |
+
- Q
|
263 |
+
- Ñ
|
264 |
+
- W
|
265 |
+
- J
|
266 |
+
- ‘
|
267 |
+
- ’
|
268 |
+
- <sos/eos>
|
269 |
+
init: xavier_uniform
|
270 |
+
input_size: null
|
271 |
+
ctc_conf:
|
272 |
+
dropout_rate: 0.0
|
273 |
+
ctc_type: builtin
|
274 |
+
reduce: true
|
275 |
+
ignore_nan_grad: null
|
276 |
+
zero_infinity: true
|
277 |
+
joint_net_conf:
|
278 |
+
joint_space_size: 640
|
279 |
+
use_preprocessor: true
|
280 |
+
token_type: bpe
|
281 |
+
bpemodel: data/eu_token_list/bpe_unigram150/bpe.model
|
282 |
+
non_linguistic_symbols: null
|
283 |
+
cleaner: null
|
284 |
+
g2p: null
|
285 |
+
speech_volume_normalize: null
|
286 |
+
rir_scp: null
|
287 |
+
rir_apply_prob: 1.0
|
288 |
+
noise_scp: null
|
289 |
+
noise_apply_prob: 1.0
|
290 |
+
noise_db_range: '13_15'
|
291 |
+
short_noise_thres: 0.5
|
292 |
+
aux_ctc_tasks: []
|
293 |
+
frontend: default
|
294 |
+
frontend_conf:
|
295 |
+
n_fft: 512
|
296 |
+
win_length: 400
|
297 |
+
hop_length: 160
|
298 |
+
fs: 16k
|
299 |
+
specaug: specaug
|
300 |
+
specaug_conf:
|
301 |
+
apply_time_warp: true
|
302 |
+
time_warp_window: 5
|
303 |
+
time_warp_mode: bicubic
|
304 |
+
apply_freq_mask: true
|
305 |
+
freq_mask_width_range:
|
306 |
+
- 0
|
307 |
+
- 30
|
308 |
+
num_freq_mask: 2
|
309 |
+
apply_time_mask: true
|
310 |
+
time_mask_width_range:
|
311 |
+
- 0
|
312 |
+
- 40
|
313 |
+
num_time_mask: 2
|
314 |
+
normalize: global_mvn
|
315 |
+
normalize_conf:
|
316 |
+
stats_file: exp/asr_stats_raw_eu_bpe150_sp/train/feats_stats.npz
|
317 |
+
model: espnet
|
318 |
+
model_conf:
|
319 |
+
ctc_weight: 0.3
|
320 |
+
lsm_weight: 0.1
|
321 |
+
length_normalized_loss: false
|
322 |
+
report_cer: true
|
323 |
+
report_wer: true
|
324 |
+
preencoder: null
|
325 |
+
preencoder_conf: {}
|
326 |
+
encoder: conformer
|
327 |
+
encoder_conf:
|
328 |
+
input_layer: conv2d
|
329 |
+
num_blocks: 12
|
330 |
+
linear_units: 2048
|
331 |
+
dropout_rate: 0.1
|
332 |
+
output_size: 256
|
333 |
+
attention_heads: 4
|
334 |
+
attention_dropout_rate: 0.0
|
335 |
+
positional_dropout_rate: 0.1
|
336 |
+
pos_enc_layer_type: rel_pos
|
337 |
+
selfattention_layer_type: rel_selfattn
|
338 |
+
activation_type: swish
|
339 |
+
rel_pos_type: latest
|
340 |
+
macaron_style: true
|
341 |
+
use_cnn_module: true
|
342 |
+
cnn_module_kernel: 15
|
343 |
+
postencoder: null
|
344 |
+
postencoder_conf: {}
|
345 |
+
decoder: transducer
|
346 |
+
decoder_conf:
|
347 |
+
rnn_type: lstm
|
348 |
+
num_layers: 1
|
349 |
+
hidden_size: 640
|
350 |
+
dropout: 0.1
|
351 |
+
dropout_embed: 0.2
|
352 |
+
preprocessor: default
|
353 |
+
preprocessor_conf: {}
|
354 |
+
required:
|
355 |
+
- output_dir
|
356 |
+
- token_list
|
357 |
+
version: '202308'
|
358 |
+
distributed: false
|
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/backward_time.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/cer_ctc.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/cer_transducer.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/clip.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/forward_time.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/gpu_max_cached_mem_GB.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/grad_norm.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/iter_time.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/loss.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/loss_ctc.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/loss_scale.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/loss_transducer.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/optim0_lr0.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/optim_step_time.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/train_time.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/images/wer_transducer.png
ADDED
![]() |
exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/valid.loss.ave_10best.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:be1538173426869ff6478b9fcd47e3d627f2dad5b4e5df92365f59d10c568bfd
|
3 |
+
size 150506813
|
exp/lm_train_lm_eu_bpe150/20epoch.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d3b426d60e27b14d0ed0656726dcaec698eb2b39ed8a0922b5cf2a55cbb9e9ed
|
3 |
+
size 27864415
|
exp/lm_train_lm_eu_bpe150/config.yaml
ADDED
@@ -0,0 +1,274 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
config: conf/train_lm.yaml
|
2 |
+
print_config: false
|
3 |
+
log_level: INFO
|
4 |
+
drop_last_iter: false
|
5 |
+
dry_run: false
|
6 |
+
iterator_type: sequence
|
7 |
+
valid_iterator_type: null
|
8 |
+
output_dir: exp/lm_train_lm_eu_bpe150
|
9 |
+
ngpu: 1
|
10 |
+
seed: 0
|
11 |
+
num_workers: 1
|
12 |
+
num_att_plot: 3
|
13 |
+
dist_backend: nccl
|
14 |
+
dist_init_method: env://
|
15 |
+
dist_world_size: null
|
16 |
+
dist_rank: null
|
17 |
+
local_rank: 0
|
18 |
+
dist_master_addr: null
|
19 |
+
dist_master_port: null
|
20 |
+
dist_launcher: null
|
21 |
+
multiprocessing_distributed: false
|
22 |
+
unused_parameters: false
|
23 |
+
sharded_ddp: false
|
24 |
+
cudnn_enabled: true
|
25 |
+
cudnn_benchmark: false
|
26 |
+
cudnn_deterministic: true
|
27 |
+
collect_stats: false
|
28 |
+
write_collected_feats: false
|
29 |
+
max_epoch: 20
|
30 |
+
patience: 3
|
31 |
+
val_scheduler_criterion:
|
32 |
+
- valid
|
33 |
+
- loss
|
34 |
+
early_stopping_criterion:
|
35 |
+
- valid
|
36 |
+
- loss
|
37 |
+
- min
|
38 |
+
best_model_criterion:
|
39 |
+
- - valid
|
40 |
+
- loss
|
41 |
+
- min
|
42 |
+
keep_nbest_models: 1
|
43 |
+
nbest_averaging_interval: 0
|
44 |
+
grad_clip: 5.0
|
45 |
+
grad_clip_type: 2.0
|
46 |
+
grad_noise: false
|
47 |
+
accum_grad: 1
|
48 |
+
no_forward_run: false
|
49 |
+
resume: true
|
50 |
+
train_dtype: float32
|
51 |
+
use_amp: false
|
52 |
+
log_interval: null
|
53 |
+
use_matplotlib: true
|
54 |
+
use_tensorboard: true
|
55 |
+
create_graph_in_tensorboard: false
|
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: 64
|
69 |
+
valid_batch_size: null
|
70 |
+
batch_bins: 1000000
|
71 |
+
valid_batch_bins: null
|
72 |
+
train_shape_file:
|
73 |
+
- exp/lm_stats_eu_bpe150/train/text_shape.bpe
|
74 |
+
valid_shape_file:
|
75 |
+
- exp/lm_stats_eu_bpe150/valid/text_shape.bpe
|
76 |
+
batch_type: folded
|
77 |
+
valid_batch_type: null
|
78 |
+
fold_length:
|
79 |
+
- 150
|
80 |
+
sort_in_batch: descending
|
81 |
+
shuffle_within_batch: false
|
82 |
+
sort_batch: descending
|
83 |
+
multiple_iterator: false
|
84 |
+
chunk_length: 500
|
85 |
+
chunk_shift_ratio: 0.5
|
86 |
+
num_cache_chunks: 1024
|
87 |
+
chunk_excluded_key_prefixes: []
|
88 |
+
train_data_path_and_name_and_type:
|
89 |
+
- - dump/raw/lm_train.txt
|
90 |
+
- text
|
91 |
+
- text
|
92 |
+
valid_data_path_and_name_and_type:
|
93 |
+
- - dump/raw/org/dev_eu/text
|
94 |
+
- text
|
95 |
+
- text
|
96 |
+
allow_variable_data_keys: false
|
97 |
+
max_cache_size: 0.0
|
98 |
+
max_cache_fd: 32
|
99 |
+
valid_max_cache_size: null
|
100 |
+
exclude_weight_decay: false
|
101 |
+
exclude_weight_decay_conf: {}
|
102 |
+
optim: sgd
|
103 |
+
optim_conf: {}
|
104 |
+
scheduler: null
|
105 |
+
scheduler_conf: {}
|
106 |
+
token_list:
|
107 |
+
- <blank>
|
108 |
+
- <unk>
|
109 |
+
- A
|
110 |
+
- ▁
|
111 |
+
- I
|
112 |
+
- E
|
113 |
+
- Z
|
114 |
+
- .
|
115 |
+
- R
|
116 |
+
- N
|
117 |
+
- U
|
118 |
+
- S
|
119 |
+
- O
|
120 |
+
- T
|
121 |
+
- KO
|
122 |
+
- K
|
123 |
+
- ▁E
|
124 |
+
- TU
|
125 |
+
- TE
|
126 |
+
- RA
|
127 |
+
- EN
|
128 |
+
- L
|
129 |
+
- ','
|
130 |
+
- LA
|
131 |
+
- TA
|
132 |
+
- AK
|
133 |
+
- ▁A
|
134 |
+
- AN
|
135 |
+
- ▁DA
|
136 |
+
- RE
|
137 |
+
- KA
|
138 |
+
- P
|
139 |
+
- GO
|
140 |
+
- IN
|
141 |
+
- B
|
142 |
+
- M
|
143 |
+
- ▁DU
|
144 |
+
- RI
|
145 |
+
- GU
|
146 |
+
- ▁ETA
|
147 |
+
- D
|
148 |
+
- ER
|
149 |
+
- UR
|
150 |
+
- ▁BA
|
151 |
+
- ▁P
|
152 |
+
- H
|
153 |
+
- MA
|
154 |
+
- ▁G
|
155 |
+
- ▁I
|
156 |
+
- ▁HA
|
157 |
+
- TZEN
|
158 |
+
- LE
|
159 |
+
- ▁EZ
|
160 |
+
- ▁O
|
161 |
+
- EK
|
162 |
+
- GI
|
163 |
+
- ▁BAT
|
164 |
+
- DA
|
165 |
+
- DU
|
166 |
+
- TZA
|
167 |
+
- KI
|
168 |
+
- DI
|
169 |
+
- RO
|
170 |
+
- ▁GA
|
171 |
+
- REN
|
172 |
+
- AR
|
173 |
+
- TEN
|
174 |
+
- GA
|
175 |
+
- TIK
|
176 |
+
- RRI
|
177 |
+
- ▁BI
|
178 |
+
- LI
|
179 |
+
- ▁BER
|
180 |
+
- G
|
181 |
+
- ▁AR
|
182 |
+
- TO
|
183 |
+
- ERA
|
184 |
+
- AREN
|
185 |
+
- ▁ZI
|
186 |
+
- ▁DE
|
187 |
+
- ▁BE
|
188 |
+
- X
|
189 |
+
- BA
|
190 |
+
- ▁DI
|
191 |
+
- ▁IZAN
|
192 |
+
- ▁ZE
|
193 |
+
- ETAN
|
194 |
+
- ▁ZEN
|
195 |
+
- EAN
|
196 |
+
- IA
|
197 |
+
- ▁JA
|
198 |
+
- ▁ERE
|
199 |
+
- ▁DITU
|
200 |
+
- ▁ZA
|
201 |
+
- ▁ERA
|
202 |
+
- LO
|
203 |
+
- ▁HOR
|
204 |
+
- NTZ
|
205 |
+
- ▁DIRA
|
206 |
+
- MEN
|
207 |
+
- ▁HI
|
208 |
+
- ▁F
|
209 |
+
- F
|
210 |
+
- LDE
|
211 |
+
- ZIO
|
212 |
+
- '?'
|
213 |
+
- ▁ZU
|
214 |
+
- '-'
|
215 |
+
- DO
|
216 |
+
- ▁EGIN
|
217 |
+
- TZEKO
|
218 |
+
- ▁BEHAR
|
219 |
+
- TZI
|
220 |
+
- BIL
|
221 |
+
- ▁IN
|
222 |
+
- RIK
|
223 |
+
- ▁HORI
|
224 |
+
- ▁SA
|
225 |
+
- ▁NA
|
226 |
+
- BIDE
|
227 |
+
- ▁KON
|
228 |
+
- ▁HE
|
229 |
+
- ▁ZUEN
|
230 |
+
- ▁MU
|
231 |
+
- ▁BESTE
|
232 |
+
- ▁SO
|
233 |
+
- ▁HERRI
|
234 |
+
- ▁IKAS
|
235 |
+
- ▁NO
|
236 |
+
- ▁ALD
|
237 |
+
- ▁NI
|
238 |
+
- ▁TX
|
239 |
+
- ABE
|
240 |
+
- KETA
|
241 |
+
- ▁BAINA
|
242 |
+
- C
|
243 |
+
- '!'
|
244 |
+
- V
|
245 |
+
- Y
|
246 |
+
- ':'
|
247 |
+
- ;
|
248 |
+
- '"'
|
249 |
+
- Í
|
250 |
+
- Q
|
251 |
+
- Ñ
|
252 |
+
- W
|
253 |
+
- J
|
254 |
+
- ‘
|
255 |
+
- ’
|
256 |
+
- <sos/eos>
|
257 |
+
init: null
|
258 |
+
model_conf:
|
259 |
+
ignore_id: 0
|
260 |
+
use_preprocessor: true
|
261 |
+
token_type: bpe
|
262 |
+
bpemodel: data/eu_token_list/bpe_unigram150/bpe.model
|
263 |
+
non_linguistic_symbols: null
|
264 |
+
cleaner: null
|
265 |
+
g2p: null
|
266 |
+
lm: seq_rnn
|
267 |
+
lm_conf:
|
268 |
+
nlayers: 2
|
269 |
+
unit: 650
|
270 |
+
required:
|
271 |
+
- output_dir
|
272 |
+
- token_list
|
273 |
+
version: '202308'
|
274 |
+
distributed: false
|
exp/lm_train_lm_eu_bpe150/images/backward_time.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/clip.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/forward_time.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/gpu_max_cached_mem_GB.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/grad_norm.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/iter_time.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/loss.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/loss_scale.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/optim0_lr0.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/optim_step_time.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/images/train_time.png
ADDED
![]() |
exp/lm_train_lm_eu_bpe150/perplexity_test/ppl
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
5.541587046132345
|
meta.yaml
ADDED
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
espnet: '202308'
|
2 |
+
files:
|
3 |
+
asr_model_file: exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/valid.loss.ave_10best.pth
|
4 |
+
lm_file: exp/lm_train_lm_eu_bpe150/20epoch.pth
|
5 |
+
python: "3.8.17 | packaged by conda-forge | (default, Jun 16 2023, 07:06:00) \n[GCC\
|
6 |
+
\ 11.4.0]"
|
7 |
+
timestamp: 1698236324.507222
|
8 |
+
torch: 2.0.1
|
9 |
+
yaml_files:
|
10 |
+
asr_train_config: exp/asr_train_asr_transducer_conformer5_raw_eu_bpe150_sp/config.yaml
|
11 |
+
lm_train_config: exp/lm_train_lm_eu_bpe150/config.yaml
|