AbeShinzo0708
commited on
Commit
•
1e51ece
1
Parent(s):
7123498
Upload 2 files
Browse files- 100epoch.pth +3 -0
- config.yaml +401 -0
100epoch.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b393b091ed1da1ef2bd4ec1d067aecc94ed7a19b4c00d4ea16d0d3ff4d233471
|
3 |
+
size 373275392
|
config.yaml
ADDED
@@ -0,0 +1,401 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
config: ./conf/tuning/finetune_full_band_vits.yaml
|
2 |
+
print_config: false
|
3 |
+
log_level: INFO
|
4 |
+
dry_run: false
|
5 |
+
iterator_type: sequence
|
6 |
+
output_dir: exp/tts_full_band_vits
|
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: 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: 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: 100
|
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 |
+
- - train
|
38 |
+
- total_count
|
39 |
+
- max
|
40 |
+
keep_nbest_models: 10
|
41 |
+
nbest_averaging_interval: 0
|
42 |
+
grad_clip: -1
|
43 |
+
grad_clip_type: 2.0
|
44 |
+
grad_noise: false
|
45 |
+
accum_grad: 1
|
46 |
+
no_forward_run: false
|
47 |
+
resume: true
|
48 |
+
train_dtype: float32
|
49 |
+
use_amp: false
|
50 |
+
log_interval: 50
|
51 |
+
use_matplotlib: true
|
52 |
+
use_tensorboard: true
|
53 |
+
create_graph_in_tensorboard: false
|
54 |
+
use_wandb: false
|
55 |
+
wandb_project: null
|
56 |
+
wandb_id: null
|
57 |
+
wandb_entity: null
|
58 |
+
wandb_name: null
|
59 |
+
wandb_model_log_interval: -1
|
60 |
+
detect_anomaly: false
|
61 |
+
pretrain_path: null
|
62 |
+
init_param:
|
63 |
+
- downloads/full_band_vits_accent_with_pause_pretrain/exp/tts_train_full_band_vits_raw_phn_jaconv_pyopenjtalk_accent_with_pause/train.total_count.ave_10best.pth:tts:tts
|
64 |
+
ignore_init_mismatch: false
|
65 |
+
freeze_param: []
|
66 |
+
num_iters_per_epoch: 1000
|
67 |
+
batch_size: 20
|
68 |
+
valid_batch_size: null
|
69 |
+
batch_bins: 100000
|
70 |
+
valid_batch_bins: null
|
71 |
+
train_shape_file:
|
72 |
+
- exp/tts_stats_raw_linear_spectrogram_phn_jaconv_pyopenjtalk_accent_with_pause/train/text_shape.phn
|
73 |
+
- exp/tts_stats_raw_linear_spectrogram_phn_jaconv_pyopenjtalk_accent_with_pause/train/speech_shape
|
74 |
+
valid_shape_file:
|
75 |
+
- exp/tts_stats_raw_linear_spectrogram_phn_jaconv_pyopenjtalk_accent_with_pause/valid/text_shape.phn
|
76 |
+
- exp/tts_stats_raw_linear_spectrogram_phn_jaconv_pyopenjtalk_accent_with_pause/valid/speech_shape
|
77 |
+
batch_type: numel
|
78 |
+
valid_batch_type: null
|
79 |
+
fold_length:
|
80 |
+
- 150
|
81 |
+
- 409600
|
82 |
+
sort_in_batch: descending
|
83 |
+
sort_batch: descending
|
84 |
+
multiple_iterator: false
|
85 |
+
chunk_length: 500
|
86 |
+
chunk_shift_ratio: 0.5
|
87 |
+
num_cache_chunks: 1024
|
88 |
+
chunk_excluded_key_prefixes: []
|
89 |
+
train_data_path_and_name_and_type:
|
90 |
+
- - dump/44k/raw/tr_no_dev/text
|
91 |
+
- text
|
92 |
+
- text
|
93 |
+
- - dump/44k/raw/tr_no_dev/wav.scp
|
94 |
+
- speech
|
95 |
+
- sound
|
96 |
+
valid_data_path_and_name_and_type:
|
97 |
+
- - dump/44k/raw/dev/text
|
98 |
+
- text
|
99 |
+
- text
|
100 |
+
- - dump/44k/raw/dev/wav.scp
|
101 |
+
- speech
|
102 |
+
- sound
|
103 |
+
allow_variable_data_keys: false
|
104 |
+
max_cache_size: 0.0
|
105 |
+
max_cache_fd: 32
|
106 |
+
valid_max_cache_size: null
|
107 |
+
exclude_weight_decay: false
|
108 |
+
exclude_weight_decay_conf: {}
|
109 |
+
optim: adamw
|
110 |
+
optim_conf:
|
111 |
+
lr: 0.0001
|
112 |
+
betas:
|
113 |
+
- 0.8
|
114 |
+
- 0.99
|
115 |
+
eps: 1.0e-09
|
116 |
+
weight_decay: 0.0
|
117 |
+
scheduler: exponentiallr
|
118 |
+
scheduler_conf:
|
119 |
+
gamma: 0.999875
|
120 |
+
optim2: adamw
|
121 |
+
optim2_conf:
|
122 |
+
lr: 0.0001
|
123 |
+
betas:
|
124 |
+
- 0.8
|
125 |
+
- 0.99
|
126 |
+
eps: 1.0e-09
|
127 |
+
weight_decay: 0.0
|
128 |
+
scheduler2: exponentiallr
|
129 |
+
scheduler2_conf:
|
130 |
+
gamma: 0.999875
|
131 |
+
generator_first: false
|
132 |
+
token_list:
|
133 |
+
- <blank>
|
134 |
+
- <unk>
|
135 |
+
- '1'
|
136 |
+
- '2'
|
137 |
+
- '0'
|
138 |
+
- '3'
|
139 |
+
- '4'
|
140 |
+
- '-1'
|
141 |
+
- '5'
|
142 |
+
- a
|
143 |
+
- o
|
144 |
+
- '-2'
|
145 |
+
- i
|
146 |
+
- '-3'
|
147 |
+
- u
|
148 |
+
- e
|
149 |
+
- k
|
150 |
+
- n
|
151 |
+
- t
|
152 |
+
- '6'
|
153 |
+
- r
|
154 |
+
- '-4'
|
155 |
+
- s
|
156 |
+
- N
|
157 |
+
- m
|
158 |
+
- pau
|
159 |
+
- '7'
|
160 |
+
- sh
|
161 |
+
- d
|
162 |
+
- g
|
163 |
+
- w
|
164 |
+
- '8'
|
165 |
+
- U
|
166 |
+
- '-5'
|
167 |
+
- I
|
168 |
+
- cl
|
169 |
+
- h
|
170 |
+
- y
|
171 |
+
- b
|
172 |
+
- '9'
|
173 |
+
- j
|
174 |
+
- ts
|
175 |
+
- ch
|
176 |
+
- '-6'
|
177 |
+
- z
|
178 |
+
- p
|
179 |
+
- '-7'
|
180 |
+
- f
|
181 |
+
- ky
|
182 |
+
- ry
|
183 |
+
- '-8'
|
184 |
+
- gy
|
185 |
+
- '-9'
|
186 |
+
- hy
|
187 |
+
- ny
|
188 |
+
- '-10'
|
189 |
+
- by
|
190 |
+
- my
|
191 |
+
- '-11'
|
192 |
+
- '-12'
|
193 |
+
- '-13'
|
194 |
+
- py
|
195 |
+
- '-14'
|
196 |
+
- '-15'
|
197 |
+
- v
|
198 |
+
- '10'
|
199 |
+
- '-16'
|
200 |
+
- '-17'
|
201 |
+
- '11'
|
202 |
+
- '-21'
|
203 |
+
- '-20'
|
204 |
+
- '12'
|
205 |
+
- '-19'
|
206 |
+
- '13'
|
207 |
+
- '-18'
|
208 |
+
- '14'
|
209 |
+
- dy
|
210 |
+
- '15'
|
211 |
+
- ty
|
212 |
+
- '-22'
|
213 |
+
- '16'
|
214 |
+
- '18'
|
215 |
+
- '19'
|
216 |
+
- '17'
|
217 |
+
- <sos/eos>
|
218 |
+
odim: null
|
219 |
+
model_conf: {}
|
220 |
+
use_preprocessor: true
|
221 |
+
token_type: phn
|
222 |
+
bpemodel: null
|
223 |
+
non_linguistic_symbols: null
|
224 |
+
cleaner: jaconv
|
225 |
+
g2p: pyopenjtalk_accent_with_pause
|
226 |
+
feats_extract: linear_spectrogram
|
227 |
+
feats_extract_conf:
|
228 |
+
n_fft: 2048
|
229 |
+
hop_length: 512
|
230 |
+
win_length: null
|
231 |
+
normalize: null
|
232 |
+
normalize_conf: {}
|
233 |
+
tts: vits
|
234 |
+
tts_conf:
|
235 |
+
generator_type: vits_generator
|
236 |
+
generator_params:
|
237 |
+
hidden_channels: 192
|
238 |
+
spks: -1
|
239 |
+
global_channels: -1
|
240 |
+
segment_size: 32
|
241 |
+
text_encoder_attention_heads: 2
|
242 |
+
text_encoder_ffn_expand: 4
|
243 |
+
text_encoder_blocks: 6
|
244 |
+
text_encoder_positionwise_layer_type: conv1d
|
245 |
+
text_encoder_positionwise_conv_kernel_size: 3
|
246 |
+
text_encoder_positional_encoding_layer_type: rel_pos
|
247 |
+
text_encoder_self_attention_layer_type: rel_selfattn
|
248 |
+
text_encoder_activation_type: swish
|
249 |
+
text_encoder_normalize_before: true
|
250 |
+
text_encoder_dropout_rate: 0.1
|
251 |
+
text_encoder_positional_dropout_rate: 0.0
|
252 |
+
text_encoder_attention_dropout_rate: 0.1
|
253 |
+
use_macaron_style_in_text_encoder: true
|
254 |
+
use_conformer_conv_in_text_encoder: false
|
255 |
+
text_encoder_conformer_kernel_size: -1
|
256 |
+
decoder_kernel_size: 7
|
257 |
+
decoder_channels: 512
|
258 |
+
decoder_upsample_scales:
|
259 |
+
- 8
|
260 |
+
- 8
|
261 |
+
- 2
|
262 |
+
- 2
|
263 |
+
- 2
|
264 |
+
decoder_upsample_kernel_sizes:
|
265 |
+
- 16
|
266 |
+
- 16
|
267 |
+
- 4
|
268 |
+
- 4
|
269 |
+
- 4
|
270 |
+
decoder_resblock_kernel_sizes:
|
271 |
+
- 3
|
272 |
+
- 7
|
273 |
+
- 11
|
274 |
+
decoder_resblock_dilations:
|
275 |
+
- - 1
|
276 |
+
- 3
|
277 |
+
- 5
|
278 |
+
- - 1
|
279 |
+
- 3
|
280 |
+
- 5
|
281 |
+
- - 1
|
282 |
+
- 3
|
283 |
+
- 5
|
284 |
+
use_weight_norm_in_decoder: true
|
285 |
+
posterior_encoder_kernel_size: 5
|
286 |
+
posterior_encoder_layers: 16
|
287 |
+
posterior_encoder_stacks: 1
|
288 |
+
posterior_encoder_base_dilation: 1
|
289 |
+
posterior_encoder_dropout_rate: 0.0
|
290 |
+
use_weight_norm_in_posterior_encoder: true
|
291 |
+
flow_flows: 4
|
292 |
+
flow_kernel_size: 5
|
293 |
+
flow_base_dilation: 1
|
294 |
+
flow_layers: 4
|
295 |
+
flow_dropout_rate: 0.0
|
296 |
+
use_weight_norm_in_flow: true
|
297 |
+
use_only_mean_in_flow: true
|
298 |
+
stochastic_duration_predictor_kernel_size: 3
|
299 |
+
stochastic_duration_predictor_dropout_rate: 0.5
|
300 |
+
stochastic_duration_predictor_flows: 4
|
301 |
+
stochastic_duration_predictor_dds_conv_layers: 3
|
302 |
+
vocabs: 85
|
303 |
+
aux_channels: 1025
|
304 |
+
discriminator_type: hifigan_multi_scale_multi_period_discriminator
|
305 |
+
discriminator_params:
|
306 |
+
scales: 1
|
307 |
+
scale_downsample_pooling: AvgPool1d
|
308 |
+
scale_downsample_pooling_params:
|
309 |
+
kernel_size: 4
|
310 |
+
stride: 2
|
311 |
+
padding: 2
|
312 |
+
scale_discriminator_params:
|
313 |
+
in_channels: 1
|
314 |
+
out_channels: 1
|
315 |
+
kernel_sizes:
|
316 |
+
- 15
|
317 |
+
- 41
|
318 |
+
- 5
|
319 |
+
- 3
|
320 |
+
channels: 128
|
321 |
+
max_downsample_channels: 1024
|
322 |
+
max_groups: 16
|
323 |
+
bias: true
|
324 |
+
downsample_scales:
|
325 |
+
- 2
|
326 |
+
- 2
|
327 |
+
- 4
|
328 |
+
- 4
|
329 |
+
- 1
|
330 |
+
nonlinear_activation: LeakyReLU
|
331 |
+
nonlinear_activation_params:
|
332 |
+
negative_slope: 0.1
|
333 |
+
use_weight_norm: true
|
334 |
+
use_spectral_norm: false
|
335 |
+
follow_official_norm: false
|
336 |
+
periods:
|
337 |
+
- 2
|
338 |
+
- 3
|
339 |
+
- 5
|
340 |
+
- 7
|
341 |
+
- 11
|
342 |
+
period_discriminator_params:
|
343 |
+
in_channels: 1
|
344 |
+
out_channels: 1
|
345 |
+
kernel_sizes:
|
346 |
+
- 5
|
347 |
+
- 3
|
348 |
+
channels: 32
|
349 |
+
downsample_scales:
|
350 |
+
- 3
|
351 |
+
- 3
|
352 |
+
- 3
|
353 |
+
- 3
|
354 |
+
- 1
|
355 |
+
max_downsample_channels: 1024
|
356 |
+
bias: true
|
357 |
+
nonlinear_activation: LeakyReLU
|
358 |
+
nonlinear_activation_params:
|
359 |
+
negative_slope: 0.1
|
360 |
+
use_weight_norm: true
|
361 |
+
use_spectral_norm: false
|
362 |
+
generator_adv_loss_params:
|
363 |
+
average_by_discriminators: false
|
364 |
+
loss_type: mse
|
365 |
+
discriminator_adv_loss_params:
|
366 |
+
average_by_discriminators: false
|
367 |
+
loss_type: mse
|
368 |
+
feat_match_loss_params:
|
369 |
+
average_by_discriminators: false
|
370 |
+
average_by_layers: false
|
371 |
+
include_final_outputs: true
|
372 |
+
mel_loss_params:
|
373 |
+
fs: 44100
|
374 |
+
n_fft: 2048
|
375 |
+
hop_length: 512
|
376 |
+
win_length: null
|
377 |
+
window: hann
|
378 |
+
n_mels: 80
|
379 |
+
fmin: 0
|
380 |
+
fmax: null
|
381 |
+
log_base: null
|
382 |
+
lambda_adv: 1.0
|
383 |
+
lambda_mel: 45.0
|
384 |
+
lambda_feat_match: 2.0
|
385 |
+
lambda_dur: 1.0
|
386 |
+
lambda_kl: 1.0
|
387 |
+
sampling_rate: 44100
|
388 |
+
cache_generator_outputs: true
|
389 |
+
pitch_extract: null
|
390 |
+
pitch_extract_conf: {}
|
391 |
+
pitch_normalize: null
|
392 |
+
pitch_normalize_conf: {}
|
393 |
+
energy_extract: null
|
394 |
+
energy_extract_conf: {}
|
395 |
+
energy_normalize: null
|
396 |
+
energy_normalize_conf: {}
|
397 |
+
required:
|
398 |
+
- output_dir
|
399 |
+
- token_list
|
400 |
+
version: '202301'
|
401 |
+
distributed: false
|