Hugging Face's logo

Spaces:
horiyouta
/
chinoTTS
Running

App Files Community
horiyouta commited on
Commit
9980eb9
β€’
1 Parent(s): bb356e6

202406091316

Browse files
Files changed (8) hide show
  1. model_assets/maya/config.json +124 -0
  2. model_assets/maya/maya_e100_s1200.safetensors +3 -0
  3. model_assets/maya/maya_e84_s1000.safetensors +3 -0
  4. model_assets/maya/style_vectors.npy +3 -0
  5. model_assets/megu/config.json +122 -0
  6. model_assets/megu/megu_e100_s1200.safetensors +3 -0
  7. model_assets/megu/megu_e84_s1000.safetensors +3 -0
  8. model_assets/megu/style_vectors.npy +3 -0
model_assets/maya/config.json ADDED
@@ -0,0 +1,124 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "maya",
3
+ "train": {
4
+ "log_interval": 200,
5
+ "eval_interval": 1000,
6
+ "seed": 42,
7
+ "epochs": 100,
8
+ "learning_rate": 0.0001,
9
+ "betas": [
10
+ 0.8,
11
+ 0.99
12
+ ],
13
+ "eps": 1e-09,
14
+ "batch_size": 4,
15
+ "bf16_run": false,
16
+ "fp16_run": false,
17
+ "lr_decay": 0.99996,
18
+ "segment_size": 16384,
19
+ "init_lr_ratio": 1,
20
+ "warmup_epochs": 0,
21
+ "c_mel": 45,
22
+ "c_kl": 1.0,
23
+ "c_commit": 100,
24
+ "skip_optimizer": false,
25
+ "freeze_ZH_bert": false,
26
+ "freeze_JP_bert": false,
27
+ "freeze_EN_bert": false,
28
+ "freeze_emo": false,
29
+ "freeze_style": false,
30
+ "freeze_decoder": false
31
+ },
32
+ "data": {
33
+ "use_jp_extra": true,
34
+ "training_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/maya/train.list",
35
+ "validation_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/maya/val.list",
36
+ "max_wav_value": 32768.0,
37
+ "sampling_rate": 44100,
38
+ "filter_length": 2048,
39
+ "hop_length": 512,
40
+ "win_length": 2048,
41
+ "n_mel_channels": 128,
42
+ "mel_fmin": 0.0,
43
+ "mel_fmax": null,
44
+ "add_blank": true,
45
+ "n_speakers": 1,
46
+ "cleaned_text": true,
47
+ "spk2id": {
48
+ "maya": 0
49
+ },
50
+ "num_styles": 7,
51
+ "style2id": {
52
+ "Neutral": 0,
53
+ "動揺": 1,
54
+ "呼びかけ": 2,
55
+ "η΄Ήδ»‹": 3,
56
+ "ε…ƒζ°—": 4,
57
+ "慰め": 5,
58
+ "ι€šεΈΈ": 6
59
+ }
60
+ },
61
+ "model": {
62
+ "use_spk_conditioned_encoder": true,
63
+ "use_noise_scaled_mas": true,
64
+ "use_mel_posterior_encoder": false,
65
+ "use_duration_discriminator": false,
66
+ "use_wavlm_discriminator": true,
67
+ "inter_channels": 192,
68
+ "hidden_channels": 192,
69
+ "filter_channels": 768,
70
+ "n_heads": 2,
71
+ "n_layers": 6,
72
+ "kernel_size": 3,
73
+ "p_dropout": 0.1,
74
+ "resblock": "1",
75
+ "resblock_kernel_sizes": [
76
+ 3,
77
+ 7,
78
+ 11
79
+ ],
80
+ "resblock_dilation_sizes": [
81
+ [
82
+ 1,
83
+ 3,
84
+ 5
85
+ ],
86
+ [
87
+ 1,
88
+ 3,
89
+ 5
90
+ ],
91
+ [
92
+ 1,
93
+ 3,
94
+ 5
95
+ ]
96
+ ],
97
+ "upsample_rates": [
98
+ 8,
99
+ 8,
100
+ 2,
101
+ 2,
102
+ 2
103
+ ],
104
+ "upsample_initial_channel": 512,
105
+ "upsample_kernel_sizes": [
106
+ 16,
107
+ 16,
108
+ 8,
109
+ 2,
110
+ 2
111
+ ],
112
+ "n_layers_q": 3,
113
+ "use_spectral_norm": false,
114
+ "gin_channels": 512,
115
+ "slm": {
116
+ "model": "./slm/wavlm-base-plus",
117
+ "sr": 16000,
118
+ "hidden": 768,
119
+ "nlayers": 13,
120
+ "initial_channel": 64
121
+ }
122
+ },
123
+ "version": "2.5.0-JP-Extra"
124
+ }
model_assets/maya/maya_e100_s1200.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01aae5b64fcdaf310ac9fd5158ac4b520eca6044b4255a1a90c697464654575f
3
+ size 251150980
model_assets/maya/maya_e84_s1000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89f0463672dcc268b7036762a2f5e8c3ead19f2f434cce40616a9fa0d050a80c
3
+ size 251150980
model_assets/maya/style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f552e8b69556dc02998e566fd0043e63cc64d656c487d7e9716d0ff90daceeed
3
+ size 7296
model_assets/megu/config.json ADDED
@@ -0,0 +1,122 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "megu",
3
+ "train": {
4
+ "log_interval": 200,
5
+ "eval_interval": 1000,
6
+ "seed": 42,
7
+ "epochs": 100,
8
+ "learning_rate": 0.0001,
9
+ "betas": [
10
+ 0.8,
11
+ 0.99
12
+ ],
13
+ "eps": 1e-09,
14
+ "batch_size": 4,
15
+ "bf16_run": false,
16
+ "fp16_run": false,
17
+ "lr_decay": 0.99996,
18
+ "segment_size": 16384,
19
+ "init_lr_ratio": 1,
20
+ "warmup_epochs": 0,
21
+ "c_mel": 45,
22
+ "c_kl": 1.0,
23
+ "c_commit": 100,
24
+ "skip_optimizer": false,
25
+ "freeze_ZH_bert": false,
26
+ "freeze_JP_bert": false,
27
+ "freeze_EN_bert": false,
28
+ "freeze_emo": false,
29
+ "freeze_style": false,
30
+ "freeze_decoder": false
31
+ },
32
+ "data": {
33
+ "use_jp_extra": true,
34
+ "training_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/megu/train.list",
35
+ "validation_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/megu/val.list",
36
+ "max_wav_value": 32768.0,
37
+ "sampling_rate": 44100,
38
+ "filter_length": 2048,
39
+ "hop_length": 512,
40
+ "win_length": 2048,
41
+ "n_mel_channels": 128,
42
+ "mel_fmin": 0.0,
43
+ "mel_fmax": null,
44
+ "add_blank": true,
45
+ "n_speakers": 1,
46
+ "cleaned_text": true,
47
+ "spk2id": {
48
+ "megu": 0
49
+ },
50
+ "num_styles": 5,
51
+ "style2id": {
52
+ "Neutral": 0,
53
+ "呼びかけ": 1,
54
+ "提摈": 2,
55
+ "ι€šεΈΈ": 3,
56
+ "ε…ƒζ°—": 4
57
+ }
58
+ },
59
+ "model": {
60
+ "use_spk_conditioned_encoder": true,
61
+ "use_noise_scaled_mas": true,
62
+ "use_mel_posterior_encoder": false,
63
+ "use_duration_discriminator": false,
64
+ "use_wavlm_discriminator": true,
65
+ "inter_channels": 192,
66
+ "hidden_channels": 192,
67
+ "filter_channels": 768,
68
+ "n_heads": 2,
69
+ "n_layers": 6,
70
+ "kernel_size": 3,
71
+ "p_dropout": 0.1,
72
+ "resblock": "1",
73
+ "resblock_kernel_sizes": [
74
+ 3,
75
+ 7,
76
+ 11
77
+ ],
78
+ "resblock_dilation_sizes": [
79
+ [
80
+ 1,
81
+ 3,
82
+ 5
83
+ ],
84
+ [
85
+ 1,
86
+ 3,
87
+ 5
88
+ ],
89
+ [
90
+ 1,
91
+ 3,
92
+ 5
93
+ ]
94
+ ],
95
+ "upsample_rates": [
96
+ 8,
97
+ 8,
98
+ 2,
99
+ 2,
100
+ 2
101
+ ],
102
+ "upsample_initial_channel": 512,
103
+ "upsample_kernel_sizes": [
104
+ 16,
105
+ 16,
106
+ 8,
107
+ 2,
108
+ 2
109
+ ],
110
+ "n_layers_q": 3,
111
+ "use_spectral_norm": false,
112
+ "gin_channels": 512,
113
+ "slm": {
114
+ "model": "./slm/wavlm-base-plus",
115
+ "sr": 16000,
116
+ "hidden": 768,
117
+ "nlayers": 13,
118
+ "initial_channel": 64
119
+ }
120
+ },
121
+ "version": "2.5.0-JP-Extra"
122
+ }
model_assets/megu/megu_e100_s1200.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0faf598293d561920f793eacd4e5ac782dda44c06b03ef58ed6b42016bad0bc
3
+ size 251150980
model_assets/megu/megu_e84_s1000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac337042b10d0c344f5b14341da4e81ff5aa0214473ad239384fbe6dacb685aa
3
+ size 251150980
model_assets/megu/style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af3112b9ccde7d21b0ceb7599060d8a9dcdb8748633f62058c0fb095a62362f1
3
+ size 5248