Upload 6 files
Browse files- .gitattributes +2 -0
- Models/maeda_kaori/config.json +117 -0
- Models/maeda_kaori/osaka_shizuku_e300_s5100.safetensors +3 -0
- Models/maeda_kaori/style_vectors.npy +0 -0
- Models/oonishi_aguri/config.json +117 -0
- Models/oonishi_aguri/style_vectors.npy +0 -0
- Models/oonishi_aguri/uehara_ayumu_e300_s4200.safetensors +3 -0
.gitattributes
CHANGED
@@ -16,3 +16,5 @@ Models/tsuchida_mahiro/mahiro_tsuchida_e100_s1300.safetensors filter=lfs diff=lf
|
|
16 |
Models/tsuchida_mahiro/mahiro_tsuchida_e77_s1000.safetensors filter=lfs diff=lfs merge=lfs -text
|
17 |
Models/sekine_hitomi/sekine_hitomi_e300_s1500.safetensors filter=lfs diff=lfs merge=lfs -text
|
18 |
Models/hinoshita_kaho/hinoshita_kaho_e1834_s11000.safetensors filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
16 |
Models/tsuchida_mahiro/mahiro_tsuchida_e77_s1000.safetensors filter=lfs diff=lfs merge=lfs -text
|
17 |
Models/sekine_hitomi/sekine_hitomi_e300_s1500.safetensors filter=lfs diff=lfs merge=lfs -text
|
18 |
Models/hinoshita_kaho/hinoshita_kaho_e1834_s11000.safetensors filter=lfs diff=lfs merge=lfs -text
|
19 |
+
Models/maeda_kaori/osaka_shizuku_e300_s5100.safetensors filter=lfs diff=lfs merge=lfs -text
|
20 |
+
Models/oonishi_aguri/uehara_ayumu_e300_s4200.safetensors filter=lfs diff=lfs merge=lfs -text
|
Models/maeda_kaori/config.json
ADDED
@@ -0,0 +1,117 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"train": {
|
3 |
+
"log_interval": 200,
|
4 |
+
"eval_interval": 1000,
|
5 |
+
"seed": 42,
|
6 |
+
"epochs": 300,
|
7 |
+
"learning_rate": 0.0001,
|
8 |
+
"betas": [
|
9 |
+
0.8,
|
10 |
+
0.99
|
11 |
+
],
|
12 |
+
"eps": 1e-09,
|
13 |
+
"batch_size": 4,
|
14 |
+
"bf16_run": false,
|
15 |
+
"fp16_run": false,
|
16 |
+
"lr_decay": 0.99996,
|
17 |
+
"segment_size": 16384,
|
18 |
+
"init_lr_ratio": 1,
|
19 |
+
"warmup_epochs": 0,
|
20 |
+
"c_mel": 45,
|
21 |
+
"c_kl": 1.0,
|
22 |
+
"c_commit": 100,
|
23 |
+
"skip_optimizer": false,
|
24 |
+
"freeze_ZH_bert": false,
|
25 |
+
"freeze_JP_bert": false,
|
26 |
+
"freeze_EN_bert": false,
|
27 |
+
"freeze_emo": false,
|
28 |
+
"freeze_style": false
|
29 |
+
},
|
30 |
+
"data": {
|
31 |
+
"use_jp_extra": true,
|
32 |
+
"training_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/osaka_shizuku/train.list",
|
33 |
+
"validation_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/osaka_shizuku/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": 1,
|
44 |
+
"cleaned_text": true,
|
45 |
+
"spk2id": {
|
46 |
+
"osaka_shizuku": 0
|
47 |
+
},
|
48 |
+
"num_styles": 1,
|
49 |
+
"style2id": {
|
50 |
+
"Neutral": 0
|
51 |
+
}
|
52 |
+
},
|
53 |
+
"model": {
|
54 |
+
"use_spk_conditioned_encoder": true,
|
55 |
+
"use_noise_scaled_mas": true,
|
56 |
+
"use_mel_posterior_encoder": false,
|
57 |
+
"use_duration_discriminator": false,
|
58 |
+
"use_wavlm_discriminator": true,
|
59 |
+
"inter_channels": 192,
|
60 |
+
"hidden_channels": 192,
|
61 |
+
"filter_channels": 768,
|
62 |
+
"n_heads": 2,
|
63 |
+
"n_layers": 6,
|
64 |
+
"kernel_size": 3,
|
65 |
+
"p_dropout": 0.1,
|
66 |
+
"resblock": "1",
|
67 |
+
"resblock_kernel_sizes": [
|
68 |
+
3,
|
69 |
+
7,
|
70 |
+
11
|
71 |
+
],
|
72 |
+
"resblock_dilation_sizes": [
|
73 |
+
[
|
74 |
+
1,
|
75 |
+
3,
|
76 |
+
5
|
77 |
+
],
|
78 |
+
[
|
79 |
+
1,
|
80 |
+
3,
|
81 |
+
5
|
82 |
+
],
|
83 |
+
[
|
84 |
+
1,
|
85 |
+
3,
|
86 |
+
5
|
87 |
+
]
|
88 |
+
],
|
89 |
+
"upsample_rates": [
|
90 |
+
8,
|
91 |
+
8,
|
92 |
+
2,
|
93 |
+
2,
|
94 |
+
2
|
95 |
+
],
|
96 |
+
"upsample_initial_channel": 512,
|
97 |
+
"upsample_kernel_sizes": [
|
98 |
+
16,
|
99 |
+
16,
|
100 |
+
8,
|
101 |
+
2,
|
102 |
+
2
|
103 |
+
],
|
104 |
+
"n_layers_q": 3,
|
105 |
+
"use_spectral_norm": false,
|
106 |
+
"gin_channels": 512,
|
107 |
+
"slm": {
|
108 |
+
"model": "./slm/wavlm-base-plus",
|
109 |
+
"sr": 16000,
|
110 |
+
"hidden": 768,
|
111 |
+
"nlayers": 13,
|
112 |
+
"initial_channel": 64
|
113 |
+
}
|
114 |
+
},
|
115 |
+
"version": "2.2-JP-Extra",
|
116 |
+
"model_name": "osaka_shizuku"
|
117 |
+
}
|
Models/maeda_kaori/osaka_shizuku_e300_s5100.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d8e6312e5645f5ee0bed51c25308495126affbaf32d266ec8b5058d356f0ef5b
|
3 |
+
size 251150980
|
Models/maeda_kaori/style_vectors.npy
ADDED
Binary file (1.15 kB). View file
|
|
Models/oonishi_aguri/config.json
ADDED
@@ -0,0 +1,117 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"train": {
|
3 |
+
"log_interval": 200,
|
4 |
+
"eval_interval": 1000,
|
5 |
+
"seed": 42,
|
6 |
+
"epochs": 300,
|
7 |
+
"learning_rate": 0.0001,
|
8 |
+
"betas": [
|
9 |
+
0.8,
|
10 |
+
0.99
|
11 |
+
],
|
12 |
+
"eps": 1e-09,
|
13 |
+
"batch_size": 4,
|
14 |
+
"bf16_run": false,
|
15 |
+
"fp16_run": false,
|
16 |
+
"lr_decay": 0.99996,
|
17 |
+
"segment_size": 16384,
|
18 |
+
"init_lr_ratio": 1,
|
19 |
+
"warmup_epochs": 0,
|
20 |
+
"c_mel": 45,
|
21 |
+
"c_kl": 1.0,
|
22 |
+
"c_commit": 100,
|
23 |
+
"skip_optimizer": false,
|
24 |
+
"freeze_ZH_bert": false,
|
25 |
+
"freeze_JP_bert": false,
|
26 |
+
"freeze_EN_bert": false,
|
27 |
+
"freeze_emo": false,
|
28 |
+
"freeze_style": false
|
29 |
+
},
|
30 |
+
"data": {
|
31 |
+
"use_jp_extra": true,
|
32 |
+
"training_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/uehara_ayumu/train.list",
|
33 |
+
"validation_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/uehara_ayumu/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": 1,
|
44 |
+
"cleaned_text": true,
|
45 |
+
"spk2id": {
|
46 |
+
"uehara_ayumu": 0
|
47 |
+
},
|
48 |
+
"num_styles": 1,
|
49 |
+
"style2id": {
|
50 |
+
"Neutral": 0
|
51 |
+
}
|
52 |
+
},
|
53 |
+
"model": {
|
54 |
+
"use_spk_conditioned_encoder": true,
|
55 |
+
"use_noise_scaled_mas": true,
|
56 |
+
"use_mel_posterior_encoder": false,
|
57 |
+
"use_duration_discriminator": false,
|
58 |
+
"use_wavlm_discriminator": true,
|
59 |
+
"inter_channels": 192,
|
60 |
+
"hidden_channels": 192,
|
61 |
+
"filter_channels": 768,
|
62 |
+
"n_heads": 2,
|
63 |
+
"n_layers": 6,
|
64 |
+
"kernel_size": 3,
|
65 |
+
"p_dropout": 0.1,
|
66 |
+
"resblock": "1",
|
67 |
+
"resblock_kernel_sizes": [
|
68 |
+
3,
|
69 |
+
7,
|
70 |
+
11
|
71 |
+
],
|
72 |
+
"resblock_dilation_sizes": [
|
73 |
+
[
|
74 |
+
1,
|
75 |
+
3,
|
76 |
+
5
|
77 |
+
],
|
78 |
+
[
|
79 |
+
1,
|
80 |
+
3,
|
81 |
+
5
|
82 |
+
],
|
83 |
+
[
|
84 |
+
1,
|
85 |
+
3,
|
86 |
+
5
|
87 |
+
]
|
88 |
+
],
|
89 |
+
"upsample_rates": [
|
90 |
+
8,
|
91 |
+
8,
|
92 |
+
2,
|
93 |
+
2,
|
94 |
+
2
|
95 |
+
],
|
96 |
+
"upsample_initial_channel": 512,
|
97 |
+
"upsample_kernel_sizes": [
|
98 |
+
16,
|
99 |
+
16,
|
100 |
+
8,
|
101 |
+
2,
|
102 |
+
2
|
103 |
+
],
|
104 |
+
"n_layers_q": 3,
|
105 |
+
"use_spectral_norm": false,
|
106 |
+
"gin_channels": 512,
|
107 |
+
"slm": {
|
108 |
+
"model": "./slm/wavlm-base-plus",
|
109 |
+
"sr": 16000,
|
110 |
+
"hidden": 768,
|
111 |
+
"nlayers": 13,
|
112 |
+
"initial_channel": 64
|
113 |
+
}
|
114 |
+
},
|
115 |
+
"version": "2.2-JP-Extra",
|
116 |
+
"model_name": "uehara_ayumu"
|
117 |
+
}
|
Models/oonishi_aguri/style_vectors.npy
ADDED
Binary file (1.15 kB). View file
|
|
Models/oonishi_aguri/uehara_ayumu_e300_s4200.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:912c12f417a51383d953120466daf638d14506f1c7b5cf991cc9489182413c25
|
3 |
+
size 251150980
|