Kit-Lemonfoot
commited on
Commit
•
c207e5f
1
Parent(s):
c2a37ae
Upload 4 files
Browse files
model_assets/SBV2_HoloJPTest2/SBV2_HoloJPTest2.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3638a4d0a3bac766f390d95c84d031aece5bb789d7cc974261e8fd9ff57d0f14
|
3 |
+
size 198775356
|
model_assets/SBV2_HoloJPTest2/config.json
ADDED
@@ -0,0 +1,123 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "SBV2_HoloJPTest2",
|
3 |
+
"train": {
|
4 |
+
"log_interval": 200,
|
5 |
+
"eval_interval": 1000,
|
6 |
+
"seed": 42,
|
7 |
+
"epochs": 100,
|
8 |
+
"learning_rate": 0.0002,
|
9 |
+
"betas": [
|
10 |
+
0.8,
|
11 |
+
0.99
|
12 |
+
],
|
13 |
+
"eps": 1e-09,
|
14 |
+
"batch_size": 4,
|
15 |
+
"bf16_run": false,
|
16 |
+
"lr_decay": 0.99995,
|
17 |
+
"segment_size": 16384,
|
18 |
+
"init_lr_ratio": 1,
|
19 |
+
"warmup_epochs": 0,
|
20 |
+
"c_mel": 45,
|
21 |
+
"c_kl": 1.0,
|
22 |
+
"skip_optimizer": false,
|
23 |
+
"freeze_ZH_bert": false,
|
24 |
+
"freeze_JP_bert": false,
|
25 |
+
"freeze_EN_bert": false,
|
26 |
+
"freeze_style": false,
|
27 |
+
"freeze_encoder": false,
|
28 |
+
"freeze_decoder": false
|
29 |
+
},
|
30 |
+
"data": {
|
31 |
+
"use_jp_extra": false,
|
32 |
+
"training_files": "Data\\SBV2_HoloJPTest2\\train.list",
|
33 |
+
"validation_files": "Data\\SBV2_HoloJPTest2\\val.list",
|
34 |
+
"max_wav_value": 32768.0,
|
35 |
+
"sampling_rate": 44100,
|
36 |
+
"filter_length": 2048,
|
37 |
+
"hop_length": 512,
|
38 |
+
"win_length": 2048,
|
39 |
+
"n_mel_channels": 128,
|
40 |
+
"mel_fmin": 0.0,
|
41 |
+
"mel_fmax": null,
|
42 |
+
"add_blank": true,
|
43 |
+
"n_speakers": 8,
|
44 |
+
"cleaned_text": true,
|
45 |
+
"num_styles": 9,
|
46 |
+
"style2id": {
|
47 |
+
"Neutral": 0,
|
48 |
+
"Koyori": 1,
|
49 |
+
"Chloe": 2,
|
50 |
+
"Lamy": 3,
|
51 |
+
"Aqua": 4,
|
52 |
+
"Sora": 5,
|
53 |
+
"Towa": 6,
|
54 |
+
"Suisei": 7,
|
55 |
+
"Ayame": 8
|
56 |
+
},
|
57 |
+
"spk2id": {
|
58 |
+
"TokinoSora": 0,
|
59 |
+
"HoshimachiSuisei": 1,
|
60 |
+
"MinatoAqua": 2,
|
61 |
+
"NakiriAyame": 3,
|
62 |
+
"TokoyamiTowa": 4,
|
63 |
+
"YukihanaLamy": 5,
|
64 |
+
"HakuiKoyori": 6,
|
65 |
+
"SakamataChloe": 7
|
66 |
+
}
|
67 |
+
},
|
68 |
+
"model": {
|
69 |
+
"use_spk_conditioned_encoder": true,
|
70 |
+
"use_noise_scaled_mas": true,
|
71 |
+
"use_mel_posterior_encoder": false,
|
72 |
+
"use_duration_discriminator": true,
|
73 |
+
"inter_channels": 192,
|
74 |
+
"hidden_channels": 192,
|
75 |
+
"filter_channels": 768,
|
76 |
+
"n_heads": 2,
|
77 |
+
"n_layers": 6,
|
78 |
+
"kernel_size": 3,
|
79 |
+
"p_dropout": 0.1,
|
80 |
+
"resblock": "1",
|
81 |
+
"resblock_kernel_sizes": [
|
82 |
+
3,
|
83 |
+
7,
|
84 |
+
11
|
85 |
+
],
|
86 |
+
"resblock_dilation_sizes": [
|
87 |
+
[
|
88 |
+
1,
|
89 |
+
3,
|
90 |
+
5
|
91 |
+
],
|
92 |
+
[
|
93 |
+
1,
|
94 |
+
3,
|
95 |
+
5
|
96 |
+
],
|
97 |
+
[
|
98 |
+
1,
|
99 |
+
3,
|
100 |
+
5
|
101 |
+
]
|
102 |
+
],
|
103 |
+
"upsample_rates": [
|
104 |
+
8,
|
105 |
+
8,
|
106 |
+
2,
|
107 |
+
2,
|
108 |
+
2
|
109 |
+
],
|
110 |
+
"upsample_initial_channel": 512,
|
111 |
+
"upsample_kernel_sizes": [
|
112 |
+
16,
|
113 |
+
16,
|
114 |
+
8,
|
115 |
+
2,
|
116 |
+
2
|
117 |
+
],
|
118 |
+
"n_layers_q": 3,
|
119 |
+
"use_spectral_norm": false,
|
120 |
+
"gin_channels": 256
|
121 |
+
},
|
122 |
+
"version": "2.4.1"
|
123 |
+
}
|
model_assets/SBV2_HoloJPTest2/style_vectors.npy
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:46baff316515c091c5a0c9c4b8423d470fc45b4e1d146301f49970dbf77e4cc7
|
3 |
+
size 9344
|
voicelist.json
CHANGED
@@ -162,45 +162,36 @@
|
|
162 |
"cover": "zeta.png"
|
163 |
},
|
164 |
"Sora": {
|
165 |
-
"enable":
|
166 |
"model_path": "SBV2_HoloJPTest2",
|
167 |
-
"title": "Tokino Sora",
|
168 |
"speakerid": "TokinoSora",
|
169 |
"datasetauthor": "Kit Lemonfoot",
|
170 |
"primarylang": "JP",
|
171 |
"cover": "sora.png"
|
172 |
},
|
173 |
"Suisei": {
|
174 |
-
"enable":
|
175 |
"model_path": "SBV2_HoloJPTest2",
|
176 |
-
"title": "Hoshimachi Suisei",
|
177 |
"speakerid": "HoshimachiSuisei",
|
178 |
"datasetauthor": "Kit Lemonfoot",
|
179 |
"primarylang": "JP",
|
180 |
"cover": "suisei.png"
|
181 |
},
|
182 |
-
"Matsuri": {
|
183 |
-
"enable": false,
|
184 |
-
"model_path": "SBV2_HoloJPTest2",
|
185 |
-
"title": "Natsuiro Matsuri",
|
186 |
-
"speakerid": "NatsuiroMatsuri",
|
187 |
-
"datasetauthor": "Kit Lemonfoot",
|
188 |
-
"primarylang": "JP",
|
189 |
-
"cover": "matsuri.png"
|
190 |
-
},
|
191 |
"Aqua": {
|
192 |
-
"enable":
|
193 |
"model_path": "SBV2_HoloJPTest2",
|
194 |
-
"title": "Minato Aqua",
|
195 |
"speakerid": "MinatoAqua",
|
196 |
"datasetauthor": "Kit Lemonfoot",
|
197 |
"primarylang": "JP",
|
198 |
"cover": "aqua.png"
|
199 |
},
|
200 |
"Ayame": {
|
201 |
-
"enable":
|
202 |
"model_path": "SBV2_HoloJPTest2",
|
203 |
-
"title": "Nakiri Ayame",
|
204 |
"speakerid": "NakiriAyame",
|
205 |
"datasetauthor": "Kit Lemonfoot",
|
206 |
"primarylang": "JP",
|
@@ -242,10 +233,19 @@
|
|
242 |
"primarylang": "JP",
|
243 |
"cover": "marine.png"
|
244 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
245 |
"Lamy": {
|
246 |
-
"enable":
|
247 |
"model_path": "SBV2_HoloJPTest2",
|
248 |
-
"title": "Yukihana Lamy",
|
249 |
"speakerid": "YukihanaLamy",
|
250 |
"datasetauthor": "Kit Lemonfoot",
|
251 |
"primarylang": "JP",
|
@@ -261,18 +261,18 @@
|
|
261 |
"cover": "laplus.png"
|
262 |
},
|
263 |
"Koyori": {
|
264 |
-
"enable":
|
265 |
"model_path": "SBV2_HoloJPTest2",
|
266 |
-
"title": "Hakui Koyori",
|
267 |
"speakerid": "HakuiKoyori",
|
268 |
"datasetauthor": "Kit Lemonfoot",
|
269 |
"primarylang": "JP",
|
270 |
"cover": "koyori.png"
|
271 |
},
|
272 |
"Chloe": {
|
273 |
-
"enable":
|
274 |
"model_path": "SBV2_HoloJPTest2",
|
275 |
-
"title": "Sakamata Chloe",
|
276 |
"speakerid": "SakamataChloe",
|
277 |
"datasetauthor": "Kit Lemonfoot",
|
278 |
"primarylang": "JP",
|
|
|
162 |
"cover": "zeta.png"
|
163 |
},
|
164 |
"Sora": {
|
165 |
+
"enable": true,
|
166 |
"model_path": "SBV2_HoloJPTest2",
|
167 |
+
"title": "Tokino Sora (WIP)",
|
168 |
"speakerid": "TokinoSora",
|
169 |
"datasetauthor": "Kit Lemonfoot",
|
170 |
"primarylang": "JP",
|
171 |
"cover": "sora.png"
|
172 |
},
|
173 |
"Suisei": {
|
174 |
+
"enable": true,
|
175 |
"model_path": "SBV2_HoloJPTest2",
|
176 |
+
"title": "Hoshimachi Suisei (WIP)",
|
177 |
"speakerid": "HoshimachiSuisei",
|
178 |
"datasetauthor": "Kit Lemonfoot",
|
179 |
"primarylang": "JP",
|
180 |
"cover": "suisei.png"
|
181 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
182 |
"Aqua": {
|
183 |
+
"enable": true,
|
184 |
"model_path": "SBV2_HoloJPTest2",
|
185 |
+
"title": "Minato Aqua (WIP)",
|
186 |
"speakerid": "MinatoAqua",
|
187 |
"datasetauthor": "Kit Lemonfoot",
|
188 |
"primarylang": "JP",
|
189 |
"cover": "aqua.png"
|
190 |
},
|
191 |
"Ayame": {
|
192 |
+
"enable": true,
|
193 |
"model_path": "SBV2_HoloJPTest2",
|
194 |
+
"title": "Nakiri Ayame (WIP)",
|
195 |
"speakerid": "NakiriAyame",
|
196 |
"datasetauthor": "Kit Lemonfoot",
|
197 |
"primarylang": "JP",
|
|
|
233 |
"primarylang": "JP",
|
234 |
"cover": "marine.png"
|
235 |
},
|
236 |
+
"Towa": {
|
237 |
+
"enable": true,
|
238 |
+
"model_path": "SBV2_HoloJPTest2",
|
239 |
+
"title": "Tokoyami Towa (WIP)",
|
240 |
+
"speakerid": "TokoyamiTowa",
|
241 |
+
"datasetauthor": "Kit Lemonfoot",
|
242 |
+
"primarylang": "JP",
|
243 |
+
"cover": "towa.png"
|
244 |
+
},
|
245 |
"Lamy": {
|
246 |
+
"enable": true,
|
247 |
"model_path": "SBV2_HoloJPTest2",
|
248 |
+
"title": "Yukihana Lamy (WIP)",
|
249 |
"speakerid": "YukihanaLamy",
|
250 |
"datasetauthor": "Kit Lemonfoot",
|
251 |
"primarylang": "JP",
|
|
|
261 |
"cover": "laplus.png"
|
262 |
},
|
263 |
"Koyori": {
|
264 |
+
"enable": true,
|
265 |
"model_path": "SBV2_HoloJPTest2",
|
266 |
+
"title": "Hakui Koyori (WIP)",
|
267 |
"speakerid": "HakuiKoyori",
|
268 |
"datasetauthor": "Kit Lemonfoot",
|
269 |
"primarylang": "JP",
|
270 |
"cover": "koyori.png"
|
271 |
},
|
272 |
"Chloe": {
|
273 |
+
"enable": true,
|
274 |
"model_path": "SBV2_HoloJPTest2",
|
275 |
+
"title": "Sakamata Chloe (WIP)",
|
276 |
"speakerid": "SakamataChloe",
|
277 |
"datasetauthor": "Kit Lemonfoot",
|
278 |
"primarylang": "JP",
|