Hugging Face's logo

Spaces:
Snow-White-995
/
Bert-VITS2-Shanghainese
Sleeping

App Files Community
1
Snow-White-995 commited on
Commit
321d27b
•
1 Parent(s): e49ac59

Upload config.json

Browse files
Files changed (1) hide show
  1. configs/config.json +318 -81
configs/config.json CHANGED
@@ -1,82 +1,319 @@
1
  {
2
- "train": {
3
- "log_interval": 200,
4
- "eval_interval": 1000,
5
- "seed": 42,
6
- "epochs": 10000,
7
- "learning_rate": 0.0002,
8
- "betas": [
9
- 0.8, 0.99
10
- ],
11
- "eps": 1e-09,
12
- "batch_size": 24,
13
- "fp16_run": false,
14
- "lr_decay": 0.99995,
15
- "segment_size": 16384,
16
- "init_lr_ratio": 1,
17
- "warmup_epochs": 0,
18
- "c_mel": 45,
19
- "c_kl": 1.0,
20
- "skip_optimizer": true
21
- },
22
- "data": {
23
- "training_files": "filelists/train.list",
24
- "validation_files": "filelists/val.list",
25
- "max_wav_value": 32768.0,
26
- "sampling_rate": 22050,
27
- "filter_length": 2048,
28
- "hop_length": 512,
29
- "win_length": 2048,
30
- "n_mel_channels": 128,
31
- "mel_fmin": 0.0,
32
- "mel_fmax": null,
33
- "add_blank": true,
34
- "n_speakers": 700,
35
- "cleaned_text": true,
36
- "spk2id": {
37
- "老克勒": 0,
38
- "美琳": 1,
39
- "Jane": 2,
40
- "小贝": 3,
41
- "小庄": 4,
42
- "小嘟": 5
43
- }
44
- },
45
- "model": {
46
- "use_spk_conditioned_encoder": true,
47
- "use_noise_scaled_mas": true,
48
- "use_mel_posterior_encoder": false,
49
- "use_duration_discriminator": true,
50
- "inter_channels": 192,
51
- "hidden_channels": 192,
52
- "filter_channels": 768,
53
- "n_heads": 2,
54
- "n_layers": 6,
55
- "kernel_size": 3,
56
- "p_dropout": 0.1,
57
- "resblock": "1",
58
- "resblock_kernel_sizes": [
59
- 3, 7, 11
60
- ],
61
- "resblock_dilation_sizes": [
62
- [
63
- 1, 3, 5
64
- ], [
65
- 1, 3, 5
66
- ], [
67
- 1, 3, 5
68
- ]
69
- ],
70
- "upsample_rates": [
71
- 8, 8, 2, 2, 2
72
- ],
73
- "upsample_initial_channel": 512,
74
- "upsample_kernel_sizes": [
75
- 16, 16, 8, 2, 2
76
- ],
77
- "n_layers_q": 3,
78
- "use_spectral_norm": false,
79
- "gin_channels": 256
80
- },
81
- "version": "2.0"
82
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  {
2
+ "train": {
3
+ "log_interval": 200,
4
+ "eval_interval": 1000,
5
+ "seed": 42,
6
+ "epochs": 10000,
7
+ "learning_rate": 0.0002,
8
+ "betas": [
9
+ 0.8,
10
+ 0.99
11
+ ],
12
+ "eps": 1e-09,
13
+ "batch_size": 24,
14
+ "fp16_run": false,
15
+ "lr_decay": 0.99995,
16
+ "segment_size": 16384,
17
+ "init_lr_ratio": 1,
18
+ "warmup_epochs": 0,
19
+ "c_mel": 45,
20
+ "c_kl": 1.0,
21
+ "skip_optimizer": true
22
+ },
23
+ "data": {
24
+ "training_files": "filelists/train.list",
25
+ "validation_files": "filelists/val.list",
26
+ "max_wav_value": 32768.0,
27
+ "sampling_rate": 22050,
28
+ "filter_length": 2048,
29
+ "hop_length": 512,
30
+ "win_length": 2048,
31
+ "n_mel_channels": 128,
32
+ "mel_fmin": 0.0,
33
+ "mel_fmax": null,
34
+ "add_blank": true,
35
+ "n_speakers": 700,
36
+ "cleaned_text": true,
37
+ "spk2id": {
38
+ "老克勒": 0,
39
+ "美琳": 1,
40
+ "Jane": 2,
41
+ "小贝": 3,
42
+ "SSB0005": 4,
43
+ "SSB0009": 5,
44
+ "SSB0011": 6,
45
+ "SSB0012": 7,
46
+ "SSB0016": 8,
47
+ "SSB0018": 9,
48
+ "SSB0033": 10,
49
+ "SSB0038": 11,
50
+ "SSB0043": 12,
51
+ "SSB0057": 13,
52
+ "SSB0073": 14,
53
+ "SSB0080": 15,
54
+ "SSB0112": 16,
55
+ "SSB0122": 17,
56
+ "SSB0133": 18,
57
+ "SSB0139": 19,
58
+ "SSB0145": 20,
59
+ "SSB0149": 21,
60
+ "SSB0193": 22,
61
+ "SSB0197": 23,
62
+ "SSB0200": 24,
63
+ "SSB0241": 25,
64
+ "SSB0246": 26,
65
+ "SSB0261": 27,
66
+ "SSB0267": 28,
67
+ "SSB0273": 29,
68
+ "SSB0287": 30,
69
+ "SSB0288": 31,
70
+ "SSB0299": 32,
71
+ "SSB0307": 33,
72
+ "SSB0309": 34,
73
+ "SSB0315": 35,
74
+ "SSB0316": 36,
75
+ "SSB0323": 37,
76
+ "SSB0338": 38,
77
+ "SSB0339": 39,
78
+ "SSB0341": 40,
79
+ "SSB0342": 41,
80
+ "SSB0354": 42,
81
+ "SSB0366": 43,
82
+ "SSB0375": 44,
83
+ "SSB0379": 45,
84
+ "SSB0380": 46,
85
+ "SSB0382": 47,
86
+ "SSB0385": 48,
87
+ "SSB0393": 49,
88
+ "SSB0394": 50,
89
+ "SSB0395": 51,
90
+ "SSB0407": 52,
91
+ "SSB0415": 53,
92
+ "SSB0426": 54,
93
+ "SSB0427": 55,
94
+ "SSB0434": 56,
95
+ "SSB0435": 57,
96
+ "SSB0470": 58,
97
+ "SSB0482": 59,
98
+ "SSB0502": 60,
99
+ "SSB0534": 61,
100
+ "SSB0535": 62,
101
+ "SSB0539": 63,
102
+ "SSB0544": 64,
103
+ "SSB0565": 65,
104
+ "SSB0570": 66,
105
+ "SSB0578": 67,
106
+ "SSB0588": 68,
107
+ "SSB0590": 69,
108
+ "SSB0594": 70,
109
+ "SSB0599": 71,
110
+ "SSB0601": 72,
111
+ "SSB0603": 73,
112
+ "SSB0606": 74,
113
+ "SSB0607": 75,
114
+ "SSB0609": 76,
115
+ "SSB0614": 77,
116
+ "SSB0623": 78,
117
+ "SSB0629": 79,
118
+ "SSB0631": 80,
119
+ "SSB0632": 81,
120
+ "SSB0666": 82,
121
+ "SSB0668": 83,
122
+ "SSB0671": 84,
123
+ "SSB0686": 85,
124
+ "SSB0700": 86,
125
+ "SSB0710": 87,
126
+ "SSB0720": 88,
127
+ "SSB0723": 89,
128
+ "SSB0737": 90,
129
+ "SSB0746": 91,
130
+ "SSB0748": 92,
131
+ "SSB0751": 93,
132
+ "SSB0758": 94,
133
+ "SSB0760": 95,
134
+ "SSB0762": 96,
135
+ "SSB0778": 97,
136
+ "SSB0780": 98,
137
+ "SSB0784": 99,
138
+ "SSB0786": 100,
139
+ "SSB0794": 101,
140
+ "SSB0817": 102,
141
+ "SSB0851": 103,
142
+ "SSB0863": 104,
143
+ "SSB0871": 105,
144
+ "SSB0887": 106,
145
+ "SSB0913": 107,
146
+ "SSB0915": 108,
147
+ "SSB0919": 109,
148
+ "SSB0935": 110,
149
+ "SSB0966": 111,
150
+ "SSB0987": 112,
151
+ "SSB1008": 113,
152
+ "SSB1020": 114,
153
+ "SSB1024": 115,
154
+ "SSB1050": 116,
155
+ "SSB1055": 117,
156
+ "SSB1056": 118,
157
+ "SSB1064": 119,
158
+ "SSB1072": 120,
159
+ "SSB1091": 121,
160
+ "SSB1096": 122,
161
+ "SSB1100": 123,
162
+ "SSB1108": 124,
163
+ "SSB1115": 125,
164
+ "SSB1125": 126,
165
+ "SSB1131": 127,
166
+ "SSB1136": 128,
167
+ "SSB1138": 129,
168
+ "SSB1161": 130,
169
+ "SSB1203": 131,
170
+ "SSB1204": 132,
171
+ "SSB1218": 133,
172
+ "SSB1221": 134,
173
+ "SSB1253": 135,
174
+ "SSB1320": 136,
175
+ "SSB1341": 137,
176
+ "SSB1366": 138,
177
+ "SSB1377": 139,
178
+ "SSB1383": 140,
179
+ "SSB1385": 141,
180
+ "SSB1392": 142,
181
+ "SSB1393": 143,
182
+ "SSB1408": 144,
183
+ "SSB1431": 145,
184
+ "SSB1437": 146,
185
+ "SSB1448": 147,
186
+ "SSB1555": 148,
187
+ "SSB1563": 149,
188
+ "SSB1567": 150,
189
+ "SSB1575": 151,
190
+ "SSB1585": 152,
191
+ "SSB1593": 153,
192
+ "SSB1607": 154,
193
+ "SSB1624": 155,
194
+ "SSB1625": 156,
195
+ "SSB1630": 157,
196
+ "SSB1650": 158,
197
+ "SSB1670": 159,
198
+ "SSB1684": 160,
199
+ "SSB1686": 161,
200
+ "SSB1699": 162,
201
+ "SSB1711": 163,
202
+ "SSB1759": 164,
203
+ "SSB1806": 165,
204
+ "SSB1828": 166,
205
+ "SSB1831": 167,
206
+ "SSB1832": 168,
207
+ "SSB1837": 169,
208
+ "SSB1846": 170,
209
+ "SSB1863": 171,
210
+ "SSB1878": 172,
211
+ "SSB1891": 173,
212
+ "SSB1918": 174,
213
+ "SSB1935": 175,
214
+ "SSB1939": 176,
215
+ "SSB1956": 177,
216
+ "SSB0693": 178,
217
+ "SSB0711": 179,
218
+ "SSB0716": 180,
219
+ "SSB0717": 181,
220
+ "SSB0736": 182,
221
+ "SSB0749": 183,
222
+ "SSB0809": 184,
223
+ "SSB0702": 185,
224
+ "SSB0822": 186,
225
+ "SSB1110": 187,
226
+ "SSB1215": 188,
227
+ "SSB1399": 189,
228
+ "SSB1728": 190,
229
+ "SSB0993": 191,
230
+ "SSB0997": 192,
231
+ "SSB1000": 193,
232
+ "SSB1001": 194,
233
+ "SSB1002": 195,
234
+ "SSB1126": 196,
235
+ "SSB1135": 197,
236
+ "SSB1176": 198,
237
+ "SSB1187": 199,
238
+ "SSB1197": 200,
239
+ "SSB1216": 201,
240
+ "SSB1219": 202,
241
+ "SSB1239": 203,
242
+ "SSB1274": 204,
243
+ "SSB1302": 205,
244
+ "SSB1322": 206,
245
+ "SSB1328": 207,
246
+ "SSB1340": 208,
247
+ "SSB1365": 209,
248
+ "SSB1382": 210,
249
+ "SSB1402": 211,
250
+ "SSB1452": 212,
251
+ "SSB1457": 213,
252
+ "SSB1739": 214,
253
+ "SSB1745": 215,
254
+ "SSB1781": 216,
255
+ "SSB1782": 217,
256
+ "SSB1809": 218,
257
+ "SSB1810": 219,
258
+ "SSB1872": 220,
259
+ "SSB1902": 221,
260
+ "小庄": 222,
261
+ "小嘟": 223
262
+ }
263
+ },
264
+ "model": {
265
+ "use_spk_conditioned_encoder": true,
266
+ "use_noise_scaled_mas": true,
267
+ "use_mel_posterior_encoder": false,
268
+ "use_duration_discriminator": true,
269
+ "inter_channels": 192,
270
+ "hidden_channels": 192,
271
+ "filter_channels": 768,
272
+ "n_heads": 2,
273
+ "n_layers": 6,
274
+ "kernel_size": 3,
275
+ "p_dropout": 0.1,
276
+ "resblock": "1",
277
+ "resblock_kernel_sizes": [
278
+ 3,
279
+ 7,
280
+ 11
281
+ ],
282
+ "resblock_dilation_sizes": [
283
+ [
284
+ 1,
285
+ 3,
286
+ 5
287
+ ],
288
+ [
289
+ 1,
290
+ 3,
291
+ 5
292
+ ],
293
+ [
294
+ 1,
295
+ 3,
296
+ 5
297
+ ]
298
+ ],
299
+ "upsample_rates": [
300
+ 8,
301
+ 8,
302
+ 2,
303
+ 2,
304
+ 2
305
+ ],
306
+ "upsample_initial_channel": 512,
307
+ "upsample_kernel_sizes": [
308
+ 16,
309
+ 16,
310
+ 8,
311
+ 2,
312
+ 2
313
+ ],
314
+ "n_layers_q": 3,
315
+ "use_spectral_norm": false,
316
+ "gin_channels": 256
317
+ },
318
+ "version": "2.0"
319
+ }