Upload 2 files
Browse files- config.yaml +232 -0
- pretrained.pth +3 -0
config.yaml
ADDED
@@ -0,0 +1,232 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
token_list:
|
2 |
+
- <blank>
|
3 |
+
- <unk>
|
4 |
+
- a
|
5 |
+
- o
|
6 |
+
- i
|
7 |
+
- '['
|
8 |
+
- '#'
|
9 |
+
- u
|
10 |
+
- ']'
|
11 |
+
- e
|
12 |
+
- k
|
13 |
+
- n
|
14 |
+
- t
|
15 |
+
- r
|
16 |
+
- s
|
17 |
+
- N
|
18 |
+
- m
|
19 |
+
- _
|
20 |
+
- sh
|
21 |
+
- d
|
22 |
+
- g
|
23 |
+
- ^
|
24 |
+
- $
|
25 |
+
- w
|
26 |
+
- cl
|
27 |
+
- h
|
28 |
+
- y
|
29 |
+
- b
|
30 |
+
- j
|
31 |
+
- ts
|
32 |
+
- ch
|
33 |
+
- z
|
34 |
+
- p
|
35 |
+
- f
|
36 |
+
- ky
|
37 |
+
- ry
|
38 |
+
- gy
|
39 |
+
- hy
|
40 |
+
- ny
|
41 |
+
- by
|
42 |
+
- my
|
43 |
+
- py
|
44 |
+
- v
|
45 |
+
- dy
|
46 |
+
- '?'
|
47 |
+
- ty
|
48 |
+
- <sos/eos>
|
49 |
+
odim: null
|
50 |
+
model_conf: {}
|
51 |
+
use_preprocessor: true
|
52 |
+
token_type: phn
|
53 |
+
bpemodel: null
|
54 |
+
non_linguistic_symbols: null
|
55 |
+
cleaner: jaconv
|
56 |
+
g2p: pyopenjtalk_prosody
|
57 |
+
feats_extract: linear_spectrogram
|
58 |
+
feats_extract_conf:
|
59 |
+
n_fft: 2048
|
60 |
+
hop_length: 512
|
61 |
+
win_length: null
|
62 |
+
normalize: null
|
63 |
+
normalize_conf: {}
|
64 |
+
tts: vits
|
65 |
+
tts_conf:
|
66 |
+
generator_type: vits_generator
|
67 |
+
generator_params:
|
68 |
+
hidden_channels: 192
|
69 |
+
spks: -1
|
70 |
+
global_channels: -1
|
71 |
+
segment_size: 32
|
72 |
+
text_encoder_attention_heads: 2
|
73 |
+
text_encoder_ffn_expand: 4
|
74 |
+
text_encoder_blocks: 6
|
75 |
+
text_encoder_positionwise_layer_type: conv1d
|
76 |
+
text_encoder_positionwise_conv_kernel_size: 3
|
77 |
+
text_encoder_positional_encoding_layer_type: rel_pos
|
78 |
+
text_encoder_self_attention_layer_type: rel_selfattn
|
79 |
+
text_encoder_activation_type: swish
|
80 |
+
text_encoder_normalize_before: true
|
81 |
+
text_encoder_dropout_rate: 0.1
|
82 |
+
text_encoder_positional_dropout_rate: 0.0
|
83 |
+
text_encoder_attention_dropout_rate: 0.1
|
84 |
+
use_macaron_style_in_text_encoder: true
|
85 |
+
use_conformer_conv_in_text_encoder: false
|
86 |
+
text_encoder_conformer_kernel_size: -1
|
87 |
+
decoder_kernel_size: 7
|
88 |
+
decoder_channels: 512
|
89 |
+
decoder_upsample_scales:
|
90 |
+
- 8
|
91 |
+
- 8
|
92 |
+
- 2
|
93 |
+
- 2
|
94 |
+
- 2
|
95 |
+
decoder_upsample_kernel_sizes:
|
96 |
+
- 16
|
97 |
+
- 16
|
98 |
+
- 4
|
99 |
+
- 4
|
100 |
+
- 4
|
101 |
+
decoder_resblock_kernel_sizes:
|
102 |
+
- 3
|
103 |
+
- 7
|
104 |
+
- 11
|
105 |
+
decoder_resblock_dilations:
|
106 |
+
- - 1
|
107 |
+
- 3
|
108 |
+
- 5
|
109 |
+
- - 1
|
110 |
+
- 3
|
111 |
+
- 5
|
112 |
+
- - 1
|
113 |
+
- 3
|
114 |
+
- 5
|
115 |
+
use_weight_norm_in_decoder: true
|
116 |
+
posterior_encoder_kernel_size: 5
|
117 |
+
posterior_encoder_layers: 16
|
118 |
+
posterior_encoder_stacks: 1
|
119 |
+
posterior_encoder_base_dilation: 1
|
120 |
+
posterior_encoder_dropout_rate: 0.0
|
121 |
+
use_weight_norm_in_posterior_encoder: true
|
122 |
+
flow_flows: 4
|
123 |
+
flow_kernel_size: 5
|
124 |
+
flow_base_dilation: 1
|
125 |
+
flow_layers: 4
|
126 |
+
flow_dropout_rate: 0.0
|
127 |
+
use_weight_norm_in_flow: true
|
128 |
+
use_only_mean_in_flow: true
|
129 |
+
stochastic_duration_predictor_kernel_size: 3
|
130 |
+
stochastic_duration_predictor_dropout_rate: 0.5
|
131 |
+
stochastic_duration_predictor_flows: 4
|
132 |
+
stochastic_duration_predictor_dds_conv_layers: 3
|
133 |
+
vocabs: 47
|
134 |
+
aux_channels: 1025
|
135 |
+
discriminator_type: hifigan_multi_scale_multi_period_discriminator
|
136 |
+
discriminator_params:
|
137 |
+
scales: 1
|
138 |
+
scale_downsample_pooling: AvgPool1d
|
139 |
+
scale_downsample_pooling_params:
|
140 |
+
kernel_size: 4
|
141 |
+
stride: 2
|
142 |
+
padding: 2
|
143 |
+
scale_discriminator_params:
|
144 |
+
in_channels: 1
|
145 |
+
out_channels: 1
|
146 |
+
kernel_sizes:
|
147 |
+
- 15
|
148 |
+
- 41
|
149 |
+
- 5
|
150 |
+
- 3
|
151 |
+
channels: 128
|
152 |
+
max_downsample_channels: 1024
|
153 |
+
max_groups: 16
|
154 |
+
bias: true
|
155 |
+
downsample_scales:
|
156 |
+
- 2
|
157 |
+
- 2
|
158 |
+
- 4
|
159 |
+
- 4
|
160 |
+
- 1
|
161 |
+
nonlinear_activation: LeakyReLU
|
162 |
+
nonlinear_activation_params:
|
163 |
+
negative_slope: 0.1
|
164 |
+
use_weight_norm: true
|
165 |
+
use_spectral_norm: false
|
166 |
+
follow_official_norm: false
|
167 |
+
periods:
|
168 |
+
- 2
|
169 |
+
- 3
|
170 |
+
- 5
|
171 |
+
- 7
|
172 |
+
- 11
|
173 |
+
period_discriminator_params:
|
174 |
+
in_channels: 1
|
175 |
+
out_channels: 1
|
176 |
+
kernel_sizes:
|
177 |
+
- 5
|
178 |
+
- 3
|
179 |
+
channels: 32
|
180 |
+
downsample_scales:
|
181 |
+
- 3
|
182 |
+
- 3
|
183 |
+
- 3
|
184 |
+
- 3
|
185 |
+
- 1
|
186 |
+
max_downsample_channels: 1024
|
187 |
+
bias: true
|
188 |
+
nonlinear_activation: LeakyReLU
|
189 |
+
nonlinear_activation_params:
|
190 |
+
negative_slope: 0.1
|
191 |
+
use_weight_norm: true
|
192 |
+
use_spectral_norm: false
|
193 |
+
generator_adv_loss_params:
|
194 |
+
average_by_discriminators: false
|
195 |
+
loss_type: mse
|
196 |
+
discriminator_adv_loss_params:
|
197 |
+
average_by_discriminators: false
|
198 |
+
loss_type: mse
|
199 |
+
feat_match_loss_params:
|
200 |
+
average_by_discriminators: false
|
201 |
+
average_by_layers: false
|
202 |
+
include_final_outputs: true
|
203 |
+
mel_loss_params:
|
204 |
+
fs: 44100
|
205 |
+
n_fft: 2048
|
206 |
+
hop_length: 512
|
207 |
+
win_length: null
|
208 |
+
window: hann
|
209 |
+
n_mels: 80
|
210 |
+
fmin: 0
|
211 |
+
fmax: null
|
212 |
+
log_base: null
|
213 |
+
lambda_adv: 1.0
|
214 |
+
lambda_mel: 45.0
|
215 |
+
lambda_feat_match: 2.0
|
216 |
+
lambda_dur: 1.0
|
217 |
+
lambda_kl: 1.0
|
218 |
+
sampling_rate: 44100
|
219 |
+
cache_generator_outputs: true
|
220 |
+
pitch_extract: null
|
221 |
+
pitch_extract_conf: {}
|
222 |
+
pitch_normalize: null
|
223 |
+
pitch_normalize_conf: {}
|
224 |
+
energy_extract: null
|
225 |
+
energy_extract_conf: {}
|
226 |
+
energy_normalize: null
|
227 |
+
energy_normalize_conf: {}
|
228 |
+
required:
|
229 |
+
- output_dir
|
230 |
+
- token_list
|
231 |
+
version: '202308'
|
232 |
+
distributed: false
|
pretrained.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:604d4f4cc757cc07c2402080ae0bdf80b473be30d278d32233514f6e28f685ac
|
3 |
+
size 373149238
|