satoshi.2020
commited on
Commit
•
b059fd8
1
Parent(s):
de4005f
Update model
Browse files- README.md +508 -0
- exp/tts_stats_raw_phn_null_g2pk/train/energy_stats.npz +0 -0
- exp/tts_stats_raw_phn_null_g2pk/train/feats_stats.npz +0 -0
- exp/tts_stats_raw_phn_null_g2pk/train/pitch_stats.npz +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/config.yaml +433 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_backward_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_fake_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_forward_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_optim_step_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_real_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_train_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_align_bin_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_align_forwardsum_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_align_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_backward_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_forward_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_g_adv_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_g_feat_match_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_g_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_g_mel_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_optim_step_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_train_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_var_dur_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_var_energy_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_var_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/generator_var_pitch_loss.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/gpu_max_cached_mem_GB.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/iter_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/optim0_lr0.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/optim1_lr0.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/images/train_time.png +0 -0
- exp/tts_train_jets_raw_phn_null_g2pk/train.total_count.ave_5best.pth +3 -0
- meta.yaml +8 -0
README.md
CHANGED
@@ -1,3 +1,511 @@
|
|
1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: cc-by-4.0
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
tags:
|
3 |
+
- espnet
|
4 |
+
- audio
|
5 |
+
- text-to-speech
|
6 |
+
language: ko
|
7 |
+
datasets:
|
8 |
+
- kss
|
9 |
license: cc-by-4.0
|
10 |
---
|
11 |
+
|
12 |
+
## ESPnet2 TTS model
|
13 |
+
|
14 |
+
### `imdanboy/kss_tts_train_jets_raw_phn_null_g2pk_train.total_count.ave`
|
15 |
+
|
16 |
+
This model was trained by satoshi.2020 using kss 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 047d0c474c18a87c205e566948410be16787e477
|
23 |
+
pip install -e .
|
24 |
+
cd egs2/kss/tts1
|
25 |
+
./run.sh --skip_data_prep false --skip_train true --download_model imdanboy/kss_tts_train_jets_raw_phn_null_g2pk_train.total_count.ave
|
26 |
+
```
|
27 |
+
|
28 |
+
|
29 |
+
|
30 |
+
## TTS config
|
31 |
+
|
32 |
+
<details><summary>expand</summary>
|
33 |
+
|
34 |
+
```
|
35 |
+
config: conf/tuning/train_jets.yaml
|
36 |
+
print_config: false
|
37 |
+
log_level: INFO
|
38 |
+
dry_run: false
|
39 |
+
iterator_type: sequence
|
40 |
+
output_dir: exp/tts_train_jets_raw_phn_null_g2pk
|
41 |
+
ngpu: 1
|
42 |
+
seed: 777
|
43 |
+
num_workers: 4
|
44 |
+
num_att_plot: 3
|
45 |
+
dist_backend: nccl
|
46 |
+
dist_init_method: env://
|
47 |
+
dist_world_size: 4
|
48 |
+
dist_rank: 0
|
49 |
+
local_rank: 0
|
50 |
+
dist_master_addr: localhost
|
51 |
+
dist_master_port: 52809
|
52 |
+
dist_launcher: null
|
53 |
+
multiprocessing_distributed: true
|
54 |
+
unused_parameters: true
|
55 |
+
sharded_ddp: false
|
56 |
+
cudnn_enabled: true
|
57 |
+
cudnn_benchmark: false
|
58 |
+
cudnn_deterministic: false
|
59 |
+
collect_stats: false
|
60 |
+
write_collected_feats: false
|
61 |
+
max_epoch: 1000
|
62 |
+
patience: null
|
63 |
+
val_scheduler_criterion:
|
64 |
+
- valid
|
65 |
+
- loss
|
66 |
+
early_stopping_criterion:
|
67 |
+
- valid
|
68 |
+
- loss
|
69 |
+
- min
|
70 |
+
best_model_criterion:
|
71 |
+
- - valid
|
72 |
+
- text2mel_loss
|
73 |
+
- min
|
74 |
+
- - train
|
75 |
+
- text2mel_loss
|
76 |
+
- min
|
77 |
+
- - train
|
78 |
+
- total_count
|
79 |
+
- max
|
80 |
+
keep_nbest_models: 5
|
81 |
+
nbest_averaging_interval: 0
|
82 |
+
grad_clip: -1
|
83 |
+
grad_clip_type: 2.0
|
84 |
+
grad_noise: false
|
85 |
+
accum_grad: 1
|
86 |
+
no_forward_run: false
|
87 |
+
resume: true
|
88 |
+
train_dtype: float32
|
89 |
+
use_amp: false
|
90 |
+
log_interval: 50
|
91 |
+
use_matplotlib: true
|
92 |
+
use_tensorboard: true
|
93 |
+
use_wandb: false
|
94 |
+
wandb_project: null
|
95 |
+
wandb_id: null
|
96 |
+
wandb_entity: null
|
97 |
+
wandb_name: null
|
98 |
+
wandb_model_log_interval: -1
|
99 |
+
detect_anomaly: false
|
100 |
+
pretrain_path: null
|
101 |
+
init_param: []
|
102 |
+
ignore_init_mismatch: false
|
103 |
+
freeze_param: []
|
104 |
+
num_iters_per_epoch: 1000
|
105 |
+
batch_size: 20
|
106 |
+
valid_batch_size: null
|
107 |
+
batch_bins: 2000000
|
108 |
+
valid_batch_bins: null
|
109 |
+
train_shape_file:
|
110 |
+
- exp/tts_stats_raw_phn_null_g2pk/train/text_shape.phn
|
111 |
+
- exp/tts_stats_raw_phn_null_g2pk/train/speech_shape
|
112 |
+
valid_shape_file:
|
113 |
+
- exp/tts_stats_raw_phn_null_g2pk/valid/text_shape.phn
|
114 |
+
- exp/tts_stats_raw_phn_null_g2pk/valid/speech_shape
|
115 |
+
batch_type: numel
|
116 |
+
valid_batch_type: null
|
117 |
+
fold_length:
|
118 |
+
- 150
|
119 |
+
- 204800
|
120 |
+
sort_in_batch: descending
|
121 |
+
sort_batch: descending
|
122 |
+
multiple_iterator: false
|
123 |
+
chunk_length: 500
|
124 |
+
chunk_shift_ratio: 0.5
|
125 |
+
num_cache_chunks: 1024
|
126 |
+
train_data_path_and_name_and_type:
|
127 |
+
- - dump/raw/tr_no_dev/text
|
128 |
+
- text
|
129 |
+
- text
|
130 |
+
- - dump/raw/tr_no_dev/wav.scp
|
131 |
+
- speech
|
132 |
+
- sound
|
133 |
+
- - exp/tts_stats_raw_phn_null_g2pk/train/collect_feats/pitch.scp
|
134 |
+
- pitch
|
135 |
+
- npy
|
136 |
+
- - exp/tts_stats_raw_phn_null_g2pk/train/collect_feats/energy.scp
|
137 |
+
- energy
|
138 |
+
- npy
|
139 |
+
valid_data_path_and_name_and_type:
|
140 |
+
- - dump/raw/dev/text
|
141 |
+
- text
|
142 |
+
- text
|
143 |
+
- - dump/raw/dev/wav.scp
|
144 |
+
- speech
|
145 |
+
- sound
|
146 |
+
- - exp/tts_stats_raw_phn_null_g2pk/valid/collect_feats/pitch.scp
|
147 |
+
- pitch
|
148 |
+
- npy
|
149 |
+
- - exp/tts_stats_raw_phn_null_g2pk/valid/collect_feats/energy.scp
|
150 |
+
- energy
|
151 |
+
- npy
|
152 |
+
allow_variable_data_keys: false
|
153 |
+
max_cache_size: 0.0
|
154 |
+
max_cache_fd: 32
|
155 |
+
valid_max_cache_size: null
|
156 |
+
optim: adamw
|
157 |
+
optim_conf:
|
158 |
+
lr: 0.0002
|
159 |
+
betas:
|
160 |
+
- 0.8
|
161 |
+
- 0.99
|
162 |
+
eps: 1.0e-09
|
163 |
+
weight_decay: 0.0
|
164 |
+
scheduler: exponentiallr
|
165 |
+
scheduler_conf:
|
166 |
+
gamma: 0.999875
|
167 |
+
optim2: adamw
|
168 |
+
optim2_conf:
|
169 |
+
lr: 0.0002
|
170 |
+
betas:
|
171 |
+
- 0.8
|
172 |
+
- 0.99
|
173 |
+
eps: 1.0e-09
|
174 |
+
weight_decay: 0.0
|
175 |
+
scheduler2: exponentiallr
|
176 |
+
scheduler2_conf:
|
177 |
+
gamma: 0.999875
|
178 |
+
generator_first: true
|
179 |
+
token_list:
|
180 |
+
- <blank>
|
181 |
+
- <unk>
|
182 |
+
- ''
|
183 |
+
- ᅡ
|
184 |
+
- ᅵ
|
185 |
+
- ᄋ
|
186 |
+
- ᅳ
|
187 |
+
- ᄀ
|
188 |
+
- ᅥ
|
189 |
+
- ᄂ
|
190 |
+
- ᆫ
|
191 |
+
- ᄅ
|
192 |
+
- ᄌ
|
193 |
+
- ᄉ
|
194 |
+
- ᅩ
|
195 |
+
- ᆯ
|
196 |
+
- ᄆ
|
197 |
+
- .
|
198 |
+
- ᅮ
|
199 |
+
- ᄃ
|
200 |
+
- ᄒ
|
201 |
+
- ᅦ
|
202 |
+
- ᆼ
|
203 |
+
- ᅢ
|
204 |
+
- ᄇ
|
205 |
+
- ᅭ
|
206 |
+
- ᅧ
|
207 |
+
- ᄊ
|
208 |
+
- ᆷ
|
209 |
+
- ᄄ
|
210 |
+
- ᆮ
|
211 |
+
- ᄎ
|
212 |
+
- ᄁ
|
213 |
+
- ᆨ
|
214 |
+
- ᄑ
|
215 |
+
- ᄐ
|
216 |
+
- ᅪ
|
217 |
+
- ᄏ
|
218 |
+
- '?'
|
219 |
+
- ᄍ
|
220 |
+
- ᆸ
|
221 |
+
- ᅬ
|
222 |
+
- ᅣ
|
223 |
+
- ᅴ
|
224 |
+
- ᅯ
|
225 |
+
- ᅨ
|
226 |
+
- ᄈ
|
227 |
+
- ᅱ
|
228 |
+
- ᅲ
|
229 |
+
- ᅫ
|
230 |
+
- ','
|
231 |
+
- '!'
|
232 |
+
- ᅤ
|
233 |
+
- ':'
|
234 |
+
- ᅰ
|
235 |
+
- ''''
|
236 |
+
- '-'
|
237 |
+
- '"'
|
238 |
+
- /
|
239 |
+
- I
|
240 |
+
- M
|
241 |
+
- F
|
242 |
+
- E
|
243 |
+
- S
|
244 |
+
- C
|
245 |
+
- A
|
246 |
+
- B
|
247 |
+
- ㅇ
|
248 |
+
- <sos/eos>
|
249 |
+
odim: null
|
250 |
+
model_conf: {}
|
251 |
+
use_preprocessor: true
|
252 |
+
token_type: phn
|
253 |
+
bpemodel: null
|
254 |
+
non_linguistic_symbols: null
|
255 |
+
cleaner: null
|
256 |
+
g2p: g2pk
|
257 |
+
feats_extract: fbank
|
258 |
+
feats_extract_conf:
|
259 |
+
n_fft: 1024
|
260 |
+
hop_length: 256
|
261 |
+
win_length: null
|
262 |
+
fs: 24000
|
263 |
+
fmin: 0
|
264 |
+
fmax: null
|
265 |
+
n_mels: 80
|
266 |
+
normalize: global_mvn
|
267 |
+
normalize_conf:
|
268 |
+
stats_file: exp/tts_stats_raw_phn_null_g2pk/train/feats_stats.npz
|
269 |
+
tts: jets
|
270 |
+
tts_conf:
|
271 |
+
generator_type: jets_generator
|
272 |
+
generator_params:
|
273 |
+
adim: 256
|
274 |
+
aheads: 2
|
275 |
+
elayers: 4
|
276 |
+
eunits: 1024
|
277 |
+
dlayers: 4
|
278 |
+
dunits: 1024
|
279 |
+
positionwise_layer_type: conv1d
|
280 |
+
positionwise_conv_kernel_size: 3
|
281 |
+
duration_predictor_layers: 2
|
282 |
+
duration_predictor_chans: 256
|
283 |
+
duration_predictor_kernel_size: 3
|
284 |
+
use_masking: true
|
285 |
+
encoder_normalize_before: true
|
286 |
+
decoder_normalize_before: true
|
287 |
+
encoder_type: transformer
|
288 |
+
decoder_type: transformer
|
289 |
+
conformer_rel_pos_type: latest
|
290 |
+
conformer_pos_enc_layer_type: rel_pos
|
291 |
+
conformer_self_attn_layer_type: rel_selfattn
|
292 |
+
conformer_activation_type: swish
|
293 |
+
use_macaron_style_in_conformer: true
|
294 |
+
use_cnn_in_conformer: true
|
295 |
+
conformer_enc_kernel_size: 7
|
296 |
+
conformer_dec_kernel_size: 31
|
297 |
+
init_type: xavier_uniform
|
298 |
+
transformer_enc_dropout_rate: 0.2
|
299 |
+
transformer_enc_positional_dropout_rate: 0.2
|
300 |
+
transformer_enc_attn_dropout_rate: 0.2
|
301 |
+
transformer_dec_dropout_rate: 0.2
|
302 |
+
transformer_dec_positional_dropout_rate: 0.2
|
303 |
+
transformer_dec_attn_dropout_rate: 0.2
|
304 |
+
pitch_predictor_layers: 5
|
305 |
+
pitch_predictor_chans: 256
|
306 |
+
pitch_predictor_kernel_size: 5
|
307 |
+
pitch_predictor_dropout: 0.5
|
308 |
+
pitch_embed_kernel_size: 1
|
309 |
+
pitch_embed_dropout: 0.0
|
310 |
+
stop_gradient_from_pitch_predictor: true
|
311 |
+
energy_predictor_layers: 2
|
312 |
+
energy_predictor_chans: 256
|
313 |
+
energy_predictor_kernel_size: 3
|
314 |
+
energy_predictor_dropout: 0.5
|
315 |
+
energy_embed_kernel_size: 1
|
316 |
+
energy_embed_dropout: 0.0
|
317 |
+
stop_gradient_from_energy_predictor: false
|
318 |
+
generator_out_channels: 1
|
319 |
+
generator_channels: 512
|
320 |
+
generator_global_channels: -1
|
321 |
+
generator_kernel_size: 7
|
322 |
+
generator_upsample_scales:
|
323 |
+
- 8
|
324 |
+
- 8
|
325 |
+
- 2
|
326 |
+
- 2
|
327 |
+
generator_upsample_kernel_sizes:
|
328 |
+
- 16
|
329 |
+
- 16
|
330 |
+
- 4
|
331 |
+
- 4
|
332 |
+
generator_resblock_kernel_sizes:
|
333 |
+
- 3
|
334 |
+
- 7
|
335 |
+
- 11
|
336 |
+
generator_resblock_dilations:
|
337 |
+
- - 1
|
338 |
+
- 3
|
339 |
+
- 5
|
340 |
+
- - 1
|
341 |
+
- 3
|
342 |
+
- 5
|
343 |
+
- - 1
|
344 |
+
- 3
|
345 |
+
- 5
|
346 |
+
generator_use_additional_convs: true
|
347 |
+
generator_bias: true
|
348 |
+
generator_nonlinear_activation: LeakyReLU
|
349 |
+
generator_nonlinear_activation_params:
|
350 |
+
negative_slope: 0.1
|
351 |
+
generator_use_weight_norm: true
|
352 |
+
segment_size: 64
|
353 |
+
idim: 69
|
354 |
+
odim: 80
|
355 |
+
discriminator_type: hifigan_multi_scale_multi_period_discriminator
|
356 |
+
discriminator_params:
|
357 |
+
scales: 1
|
358 |
+
scale_downsample_pooling: AvgPool1d
|
359 |
+
scale_downsample_pooling_params:
|
360 |
+
kernel_size: 4
|
361 |
+
stride: 2
|
362 |
+
padding: 2
|
363 |
+
scale_discriminator_params:
|
364 |
+
in_channels: 1
|
365 |
+
out_channels: 1
|
366 |
+
kernel_sizes:
|
367 |
+
- 15
|
368 |
+
- 41
|
369 |
+
- 5
|
370 |
+
- 3
|
371 |
+
channels: 128
|
372 |
+
max_downsample_channels: 1024
|
373 |
+
max_groups: 16
|
374 |
+
bias: true
|
375 |
+
downsample_scales:
|
376 |
+
- 2
|
377 |
+
- 2
|
378 |
+
- 4
|
379 |
+
- 4
|
380 |
+
- 1
|
381 |
+
nonlinear_activation: LeakyReLU
|
382 |
+
nonlinear_activation_params:
|
383 |
+
negative_slope: 0.1
|
384 |
+
use_weight_norm: true
|
385 |
+
use_spectral_norm: false
|
386 |
+
follow_official_norm: false
|
387 |
+
periods:
|
388 |
+
- 2
|
389 |
+
- 3
|
390 |
+
- 5
|
391 |
+
- 7
|
392 |
+
- 11
|
393 |
+
period_discriminator_params:
|
394 |
+
in_channels: 1
|
395 |
+
out_channels: 1
|
396 |
+
kernel_sizes:
|
397 |
+
- 5
|
398 |
+
- 3
|
399 |
+
channels: 32
|
400 |
+
downsample_scales:
|
401 |
+
- 3
|
402 |
+
- 3
|
403 |
+
- 3
|
404 |
+
- 3
|
405 |
+
- 1
|
406 |
+
max_downsample_channels: 1024
|
407 |
+
bias: true
|
408 |
+
nonlinear_activation: LeakyReLU
|
409 |
+
nonlinear_activation_params:
|
410 |
+
negative_slope: 0.1
|
411 |
+
use_weight_norm: true
|
412 |
+
use_spectral_norm: false
|
413 |
+
generator_adv_loss_params:
|
414 |
+
average_by_discriminators: false
|
415 |
+
loss_type: mse
|
416 |
+
discriminator_adv_loss_params:
|
417 |
+
average_by_discriminators: false
|
418 |
+
loss_type: mse
|
419 |
+
feat_match_loss_params:
|
420 |
+
average_by_discriminators: false
|
421 |
+
average_by_layers: false
|
422 |
+
include_final_outputs: true
|
423 |
+
mel_loss_params:
|
424 |
+
fs: 24000
|
425 |
+
n_fft: 1024
|
426 |
+
hop_length: 256
|
427 |
+
win_length: null
|
428 |
+
window: hann
|
429 |
+
n_mels: 80
|
430 |
+
fmin: 0
|
431 |
+
fmax: null
|
432 |
+
log_base: null
|
433 |
+
lambda_adv: 1.0
|
434 |
+
lambda_mel: 45.0
|
435 |
+
lambda_feat_match: 2.0
|
436 |
+
lambda_var: 1.0
|
437 |
+
lambda_align: 2.0
|
438 |
+
sampling_rate: 24000
|
439 |
+
cache_generator_outputs: true
|
440 |
+
pitch_extract: dio
|
441 |
+
pitch_extract_conf:
|
442 |
+
reduction_factor: 1
|
443 |
+
use_token_averaged_f0: false
|
444 |
+
fs: 24000
|
445 |
+
n_fft: 1024
|
446 |
+
hop_length: 256
|
447 |
+
f0max: 400
|
448 |
+
f0min: 80
|
449 |
+
pitch_normalize: global_mvn
|
450 |
+
pitch_normalize_conf:
|
451 |
+
stats_file: exp/tts_stats_raw_phn_null_g2pk/train/pitch_stats.npz
|
452 |
+
energy_extract: energy
|
453 |
+
energy_extract_conf:
|
454 |
+
reduction_factor: 1
|
455 |
+
use_token_averaged_energy: false
|
456 |
+
fs: 24000
|
457 |
+
n_fft: 1024
|
458 |
+
hop_length: 256
|
459 |
+
win_length: null
|
460 |
+
energy_normalize: global_mvn
|
461 |
+
energy_normalize_conf:
|
462 |
+
stats_file: exp/tts_stats_raw_phn_null_g2pk/train/energy_stats.npz
|
463 |
+
required:
|
464 |
+
- output_dir
|
465 |
+
- token_list
|
466 |
+
version: '202204'
|
467 |
+
distributed: true
|
468 |
+
```
|
469 |
+
|
470 |
+
</details>
|
471 |
+
|
472 |
+
|
473 |
+
|
474 |
+
### Citing ESPnet
|
475 |
+
|
476 |
+
```BibTex
|
477 |
+
@inproceedings{watanabe2018espnet,
|
478 |
+
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},
|
479 |
+
title={{ESPnet}: End-to-End Speech Processing Toolkit},
|
480 |
+
year={2018},
|
481 |
+
booktitle={Proceedings of Interspeech},
|
482 |
+
pages={2207--2211},
|
483 |
+
doi={10.21437/Interspeech.2018-1456},
|
484 |
+
url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
|
485 |
+
}
|
486 |
+
|
487 |
+
|
488 |
+
|
489 |
+
|
490 |
+
@inproceedings{hayashi2020espnet,
|
491 |
+
title={{Espnet-TTS}: Unified, reproducible, and integratable open source end-to-end text-to-speech toolkit},
|
492 |
+
author={Hayashi, Tomoki and Yamamoto, Ryuichi and Inoue, Katsuki and Yoshimura, Takenori and Watanabe, Shinji and Toda, Tomoki and Takeda, Kazuya and Zhang, Yu and Tan, Xu},
|
493 |
+
booktitle={Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
|
494 |
+
pages={7654--7658},
|
495 |
+
year={2020},
|
496 |
+
organization={IEEE}
|
497 |
+
}
|
498 |
+
```
|
499 |
+
|
500 |
+
or arXiv:
|
501 |
+
|
502 |
+
```bibtex
|
503 |
+
@misc{watanabe2018espnet,
|
504 |
+
title={ESPnet: End-to-End Speech Processing Toolkit},
|
505 |
+
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},
|
506 |
+
year={2018},
|
507 |
+
eprint={1804.00015},
|
508 |
+
archivePrefix={arXiv},
|
509 |
+
primaryClass={cs.CL}
|
510 |
+
}
|
511 |
+
```
|
exp/tts_stats_raw_phn_null_g2pk/train/energy_stats.npz
ADDED
Binary file (770 Bytes). View file
|
|
exp/tts_stats_raw_phn_null_g2pk/train/feats_stats.npz
ADDED
Binary file (1.4 kB). View file
|
|
exp/tts_stats_raw_phn_null_g2pk/train/pitch_stats.npz
ADDED
Binary file (770 Bytes). View file
|
|
exp/tts_train_jets_raw_phn_null_g2pk/config.yaml
ADDED
@@ -0,0 +1,433 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
config: conf/tuning/train_jets.yaml
|
2 |
+
print_config: false
|
3 |
+
log_level: INFO
|
4 |
+
dry_run: false
|
5 |
+
iterator_type: sequence
|
6 |
+
output_dir: exp/tts_train_jets_raw_phn_null_g2pk
|
7 |
+
ngpu: 1
|
8 |
+
seed: 777
|
9 |
+
num_workers: 4
|
10 |
+
num_att_plot: 3
|
11 |
+
dist_backend: nccl
|
12 |
+
dist_init_method: env://
|
13 |
+
dist_world_size: 4
|
14 |
+
dist_rank: 0
|
15 |
+
local_rank: 0
|
16 |
+
dist_master_addr: localhost
|
17 |
+
dist_master_port: 52809
|
18 |
+
dist_launcher: null
|
19 |
+
multiprocessing_distributed: true
|
20 |
+
unused_parameters: true
|
21 |
+
sharded_ddp: false
|
22 |
+
cudnn_enabled: true
|
23 |
+
cudnn_benchmark: false
|
24 |
+
cudnn_deterministic: false
|
25 |
+
collect_stats: false
|
26 |
+
write_collected_feats: false
|
27 |
+
max_epoch: 1000
|
28 |
+
patience: null
|
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 |
+
- text2mel_loss
|
39 |
+
- min
|
40 |
+
- - train
|
41 |
+
- text2mel_loss
|
42 |
+
- min
|
43 |
+
- - train
|
44 |
+
- total_count
|
45 |
+
- max
|
46 |
+
keep_nbest_models: 5
|
47 |
+
nbest_averaging_interval: 0
|
48 |
+
grad_clip: -1
|
49 |
+
grad_clip_type: 2.0
|
50 |
+
grad_noise: false
|
51 |
+
accum_grad: 1
|
52 |
+
no_forward_run: false
|
53 |
+
resume: true
|
54 |
+
train_dtype: float32
|
55 |
+
use_amp: false
|
56 |
+
log_interval: 50
|
57 |
+
use_matplotlib: true
|
58 |
+
use_tensorboard: true
|
59 |
+
use_wandb: false
|
60 |
+
wandb_project: null
|
61 |
+
wandb_id: null
|
62 |
+
wandb_entity: null
|
63 |
+
wandb_name: null
|
64 |
+
wandb_model_log_interval: -1
|
65 |
+
detect_anomaly: false
|
66 |
+
pretrain_path: null
|
67 |
+
init_param: []
|
68 |
+
ignore_init_mismatch: false
|
69 |
+
freeze_param: []
|
70 |
+
num_iters_per_epoch: 1000
|
71 |
+
batch_size: 20
|
72 |
+
valid_batch_size: null
|
73 |
+
batch_bins: 2000000
|
74 |
+
valid_batch_bins: null
|
75 |
+
train_shape_file:
|
76 |
+
- exp/tts_stats_raw_phn_null_g2pk/train/text_shape.phn
|
77 |
+
- exp/tts_stats_raw_phn_null_g2pk/train/speech_shape
|
78 |
+
valid_shape_file:
|
79 |
+
- exp/tts_stats_raw_phn_null_g2pk/valid/text_shape.phn
|
80 |
+
- exp/tts_stats_raw_phn_null_g2pk/valid/speech_shape
|
81 |
+
batch_type: numel
|
82 |
+
valid_batch_type: null
|
83 |
+
fold_length:
|
84 |
+
- 150
|
85 |
+
- 204800
|
86 |
+
sort_in_batch: descending
|
87 |
+
sort_batch: descending
|
88 |
+
multiple_iterator: false
|
89 |
+
chunk_length: 500
|
90 |
+
chunk_shift_ratio: 0.5
|
91 |
+
num_cache_chunks: 1024
|
92 |
+
train_data_path_and_name_and_type:
|
93 |
+
- - dump/raw/tr_no_dev/text
|
94 |
+
- text
|
95 |
+
- text
|
96 |
+
- - dump/raw/tr_no_dev/wav.scp
|
97 |
+
- speech
|
98 |
+
- sound
|
99 |
+
- - exp/tts_stats_raw_phn_null_g2pk/train/collect_feats/pitch.scp
|
100 |
+
- pitch
|
101 |
+
- npy
|
102 |
+
- - exp/tts_stats_raw_phn_null_g2pk/train/collect_feats/energy.scp
|
103 |
+
- energy
|
104 |
+
- npy
|
105 |
+
valid_data_path_and_name_and_type:
|
106 |
+
- - dump/raw/dev/text
|
107 |
+
- text
|
108 |
+
- text
|
109 |
+
- - dump/raw/dev/wav.scp
|
110 |
+
- speech
|
111 |
+
- sound
|
112 |
+
- - exp/tts_stats_raw_phn_null_g2pk/valid/collect_feats/pitch.scp
|
113 |
+
- pitch
|
114 |
+
- npy
|
115 |
+
- - exp/tts_stats_raw_phn_null_g2pk/valid/collect_feats/energy.scp
|
116 |
+
- energy
|
117 |
+
- npy
|
118 |
+
allow_variable_data_keys: false
|
119 |
+
max_cache_size: 0.0
|
120 |
+
max_cache_fd: 32
|
121 |
+
valid_max_cache_size: null
|
122 |
+
optim: adamw
|
123 |
+
optim_conf:
|
124 |
+
lr: 0.0002
|
125 |
+
betas:
|
126 |
+
- 0.8
|
127 |
+
- 0.99
|
128 |
+
eps: 1.0e-09
|
129 |
+
weight_decay: 0.0
|
130 |
+
scheduler: exponentiallr
|
131 |
+
scheduler_conf:
|
132 |
+
gamma: 0.999875
|
133 |
+
optim2: adamw
|
134 |
+
optim2_conf:
|
135 |
+
lr: 0.0002
|
136 |
+
betas:
|
137 |
+
- 0.8
|
138 |
+
- 0.99
|
139 |
+
eps: 1.0e-09
|
140 |
+
weight_decay: 0.0
|
141 |
+
scheduler2: exponentiallr
|
142 |
+
scheduler2_conf:
|
143 |
+
gamma: 0.999875
|
144 |
+
generator_first: true
|
145 |
+
token_list:
|
146 |
+
- <blank>
|
147 |
+
- <unk>
|
148 |
+
- ''
|
149 |
+
- ᅡ
|
150 |
+
- ᅵ
|
151 |
+
- ᄋ
|
152 |
+
- ᅳ
|
153 |
+
- ᄀ
|
154 |
+
- ᅥ
|
155 |
+
- ᄂ
|
156 |
+
- ᆫ
|
157 |
+
- ᄅ
|
158 |
+
- ᄌ
|
159 |
+
- ᄉ
|
160 |
+
- ᅩ
|
161 |
+
- ᆯ
|
162 |
+
- ᄆ
|
163 |
+
- .
|
164 |
+
- ᅮ
|
165 |
+
- ᄃ
|
166 |
+
- ᄒ
|
167 |
+
- ᅦ
|
168 |
+
- ᆼ
|
169 |
+
- ᅢ
|
170 |
+
- ᄇ
|
171 |
+
- ᅭ
|
172 |
+
- ᅧ
|
173 |
+
- ᄊ
|
174 |
+
- ᆷ
|
175 |
+
- ᄄ
|
176 |
+
- ᆮ
|
177 |
+
- ᄎ
|
178 |
+
- ᄁ
|
179 |
+
- ᆨ
|
180 |
+
- ᄑ
|
181 |
+
- ᄐ
|
182 |
+
- ᅪ
|
183 |
+
- ᄏ
|
184 |
+
- '?'
|
185 |
+
- ᄍ
|
186 |
+
- ᆸ
|
187 |
+
- ᅬ
|
188 |
+
- ᅣ
|
189 |
+
- ᅴ
|
190 |
+
- ᅯ
|
191 |
+
- ᅨ
|
192 |
+
- ᄈ
|
193 |
+
- ᅱ
|
194 |
+
- ᅲ
|
195 |
+
- ᅫ
|
196 |
+
- ','
|
197 |
+
- '!'
|
198 |
+
- ᅤ
|
199 |
+
- ':'
|
200 |
+
- ᅰ
|
201 |
+
- ''''
|
202 |
+
- '-'
|
203 |
+
- '"'
|
204 |
+
- /
|
205 |
+
- I
|
206 |
+
- M
|
207 |
+
- F
|
208 |
+
- E
|
209 |
+
- S
|
210 |
+
- C
|
211 |
+
- A
|
212 |
+
- B
|
213 |
+
- ㅇ
|
214 |
+
- <sos/eos>
|
215 |
+
odim: null
|
216 |
+
model_conf: {}
|
217 |
+
use_preprocessor: true
|
218 |
+
token_type: phn
|
219 |
+
bpemodel: null
|
220 |
+
non_linguistic_symbols: null
|
221 |
+
cleaner: null
|
222 |
+
g2p: g2pk
|
223 |
+
feats_extract: fbank
|
224 |
+
feats_extract_conf:
|
225 |
+
n_fft: 1024
|
226 |
+
hop_length: 256
|
227 |
+
win_length: null
|
228 |
+
fs: 24000
|
229 |
+
fmin: 0
|
230 |
+
fmax: null
|
231 |
+
n_mels: 80
|
232 |
+
normalize: global_mvn
|
233 |
+
normalize_conf:
|
234 |
+
stats_file: exp/tts_stats_raw_phn_null_g2pk/train/feats_stats.npz
|
235 |
+
tts: jets
|
236 |
+
tts_conf:
|
237 |
+
generator_type: jets_generator
|
238 |
+
generator_params:
|
239 |
+
adim: 256
|
240 |
+
aheads: 2
|
241 |
+
elayers: 4
|
242 |
+
eunits: 1024
|
243 |
+
dlayers: 4
|
244 |
+
dunits: 1024
|
245 |
+
positionwise_layer_type: conv1d
|
246 |
+
positionwise_conv_kernel_size: 3
|
247 |
+
duration_predictor_layers: 2
|
248 |
+
duration_predictor_chans: 256
|
249 |
+
duration_predictor_kernel_size: 3
|
250 |
+
use_masking: true
|
251 |
+
encoder_normalize_before: true
|
252 |
+
decoder_normalize_before: true
|
253 |
+
encoder_type: transformer
|
254 |
+
decoder_type: transformer
|
255 |
+
conformer_rel_pos_type: latest
|
256 |
+
conformer_pos_enc_layer_type: rel_pos
|
257 |
+
conformer_self_attn_layer_type: rel_selfattn
|
258 |
+
conformer_activation_type: swish
|
259 |
+
use_macaron_style_in_conformer: true
|
260 |
+
use_cnn_in_conformer: true
|
261 |
+
conformer_enc_kernel_size: 7
|
262 |
+
conformer_dec_kernel_size: 31
|
263 |
+
init_type: xavier_uniform
|
264 |
+
transformer_enc_dropout_rate: 0.2
|
265 |
+
transformer_enc_positional_dropout_rate: 0.2
|
266 |
+
transformer_enc_attn_dropout_rate: 0.2
|
267 |
+
transformer_dec_dropout_rate: 0.2
|
268 |
+
transformer_dec_positional_dropout_rate: 0.2
|
269 |
+
transformer_dec_attn_dropout_rate: 0.2
|
270 |
+
pitch_predictor_layers: 5
|
271 |
+
pitch_predictor_chans: 256
|
272 |
+
pitch_predictor_kernel_size: 5
|
273 |
+
pitch_predictor_dropout: 0.5
|
274 |
+
pitch_embed_kernel_size: 1
|
275 |
+
pitch_embed_dropout: 0.0
|
276 |
+
stop_gradient_from_pitch_predictor: true
|
277 |
+
energy_predictor_layers: 2
|
278 |
+
energy_predictor_chans: 256
|
279 |
+
energy_predictor_kernel_size: 3
|
280 |
+
energy_predictor_dropout: 0.5
|
281 |
+
energy_embed_kernel_size: 1
|
282 |
+
energy_embed_dropout: 0.0
|
283 |
+
stop_gradient_from_energy_predictor: false
|
284 |
+
generator_out_channels: 1
|
285 |
+
generator_channels: 512
|
286 |
+
generator_global_channels: -1
|
287 |
+
generator_kernel_size: 7
|
288 |
+
generator_upsample_scales:
|
289 |
+
- 8
|
290 |
+
- 8
|
291 |
+
- 2
|
292 |
+
- 2
|
293 |
+
generator_upsample_kernel_sizes:
|
294 |
+
- 16
|
295 |
+
- 16
|
296 |
+
- 4
|
297 |
+
- 4
|
298 |
+
generator_resblock_kernel_sizes:
|
299 |
+
- 3
|
300 |
+
- 7
|
301 |
+
- 11
|
302 |
+
generator_resblock_dilations:
|
303 |
+
- - 1
|
304 |
+
- 3
|
305 |
+
- 5
|
306 |
+
- - 1
|
307 |
+
- 3
|
308 |
+
- 5
|
309 |
+
- - 1
|
310 |
+
- 3
|
311 |
+
- 5
|
312 |
+
generator_use_additional_convs: true
|
313 |
+
generator_bias: true
|
314 |
+
generator_nonlinear_activation: LeakyReLU
|
315 |
+
generator_nonlinear_activation_params:
|
316 |
+
negative_slope: 0.1
|
317 |
+
generator_use_weight_norm: true
|
318 |
+
segment_size: 64
|
319 |
+
idim: 69
|
320 |
+
odim: 80
|
321 |
+
discriminator_type: hifigan_multi_scale_multi_period_discriminator
|
322 |
+
discriminator_params:
|
323 |
+
scales: 1
|
324 |
+
scale_downsample_pooling: AvgPool1d
|
325 |
+
scale_downsample_pooling_params:
|
326 |
+
kernel_size: 4
|
327 |
+
stride: 2
|
328 |
+
padding: 2
|
329 |
+
scale_discriminator_params:
|
330 |
+
in_channels: 1
|
331 |
+
out_channels: 1
|
332 |
+
kernel_sizes:
|
333 |
+
- 15
|
334 |
+
- 41
|
335 |
+
- 5
|
336 |
+
- 3
|
337 |
+
channels: 128
|
338 |
+
max_downsample_channels: 1024
|
339 |
+
max_groups: 16
|
340 |
+
bias: true
|
341 |
+
downsample_scales:
|
342 |
+
- 2
|
343 |
+
- 2
|
344 |
+
- 4
|
345 |
+
- 4
|
346 |
+
- 1
|
347 |
+
nonlinear_activation: LeakyReLU
|
348 |
+
nonlinear_activation_params:
|
349 |
+
negative_slope: 0.1
|
350 |
+
use_weight_norm: true
|
351 |
+
use_spectral_norm: false
|
352 |
+
follow_official_norm: false
|
353 |
+
periods:
|
354 |
+
- 2
|
355 |
+
- 3
|
356 |
+
- 5
|
357 |
+
- 7
|
358 |
+
- 11
|
359 |
+
period_discriminator_params:
|
360 |
+
in_channels: 1
|
361 |
+
out_channels: 1
|
362 |
+
kernel_sizes:
|
363 |
+
- 5
|
364 |
+
- 3
|
365 |
+
channels: 32
|
366 |
+
downsample_scales:
|
367 |
+
- 3
|
368 |
+
- 3
|
369 |
+
- 3
|
370 |
+
- 3
|
371 |
+
- 1
|
372 |
+
max_downsample_channels: 1024
|
373 |
+
bias: true
|
374 |
+
nonlinear_activation: LeakyReLU
|
375 |
+
nonlinear_activation_params:
|
376 |
+
negative_slope: 0.1
|
377 |
+
use_weight_norm: true
|
378 |
+
use_spectral_norm: false
|
379 |
+
generator_adv_loss_params:
|
380 |
+
average_by_discriminators: false
|
381 |
+
loss_type: mse
|
382 |
+
discriminator_adv_loss_params:
|
383 |
+
average_by_discriminators: false
|
384 |
+
loss_type: mse
|
385 |
+
feat_match_loss_params:
|
386 |
+
average_by_discriminators: false
|
387 |
+
average_by_layers: false
|
388 |
+
include_final_outputs: true
|
389 |
+
mel_loss_params:
|
390 |
+
fs: 24000
|
391 |
+
n_fft: 1024
|
392 |
+
hop_length: 256
|
393 |
+
win_length: null
|
394 |
+
window: hann
|
395 |
+
n_mels: 80
|
396 |
+
fmin: 0
|
397 |
+
fmax: null
|
398 |
+
log_base: null
|
399 |
+
lambda_adv: 1.0
|
400 |
+
lambda_mel: 45.0
|
401 |
+
lambda_feat_match: 2.0
|
402 |
+
lambda_var: 1.0
|
403 |
+
lambda_align: 2.0
|
404 |
+
sampling_rate: 24000
|
405 |
+
cache_generator_outputs: true
|
406 |
+
pitch_extract: dio
|
407 |
+
pitch_extract_conf:
|
408 |
+
reduction_factor: 1
|
409 |
+
use_token_averaged_f0: false
|
410 |
+
fs: 24000
|
411 |
+
n_fft: 1024
|
412 |
+
hop_length: 256
|
413 |
+
f0max: 400
|
414 |
+
f0min: 80
|
415 |
+
pitch_normalize: global_mvn
|
416 |
+
pitch_normalize_conf:
|
417 |
+
stats_file: exp/tts_stats_raw_phn_null_g2pk/train/pitch_stats.npz
|
418 |
+
energy_extract: energy
|
419 |
+
energy_extract_conf:
|
420 |
+
reduction_factor: 1
|
421 |
+
use_token_averaged_energy: false
|
422 |
+
fs: 24000
|
423 |
+
n_fft: 1024
|
424 |
+
hop_length: 256
|
425 |
+
win_length: null
|
426 |
+
energy_normalize: global_mvn
|
427 |
+
energy_normalize_conf:
|
428 |
+
stats_file: exp/tts_stats_raw_phn_null_g2pk/train/energy_stats.npz
|
429 |
+
required:
|
430 |
+
- output_dir
|
431 |
+
- token_list
|
432 |
+
version: '202204'
|
433 |
+
distributed: true
|
exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_backward_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_fake_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_forward_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_optim_step_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_real_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/discriminator_train_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_align_bin_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_align_forwardsum_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_align_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_backward_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_forward_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_g_adv_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_g_feat_match_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_g_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_g_mel_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_optim_step_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_train_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_var_dur_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_var_energy_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_var_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/generator_var_pitch_loss.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/gpu_max_cached_mem_GB.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/iter_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/optim0_lr0.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/optim1_lr0.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/images/train_time.png
ADDED
exp/tts_train_jets_raw_phn_null_g2pk/train.total_count.ave_5best.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5de866278f02f5993f647031c4a822363a64fd23a9847a417ad18da71f479b2d
|
3 |
+
size 333680139
|
meta.yaml
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
espnet: '202204'
|
2 |
+
files:
|
3 |
+
model_file: exp/tts_train_jets_raw_phn_null_g2pk/train.total_count.ave_5best.pth
|
4 |
+
python: "3.7.11 (default, Jul 27 2021, 14:32:16) \n[GCC 7.5.0]"
|
5 |
+
timestamp: 1653840124.881467
|
6 |
+
torch: 1.10.1+cu113
|
7 |
+
yaml_files:
|
8 |
+
train_config: exp/tts_train_jets_raw_phn_null_g2pk/config.yaml
|