horiyouta commited on
Commit
bb356e6
Β·
1 Parent(s): 215d7ce

202406081608

Browse files
model_assets/.gitignore CHANGED
@@ -1,2 +1,2 @@
1
- *
2
  !.gitignore
 
1
+ # *
2
  !.gitignore
model_assets/rize/config.json ADDED
@@ -0,0 +1,123 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "rize",
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/rize/train.list",
35
+ "validation_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/rize/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
+ "rize": 0
49
+ },
50
+ "num_styles": 6,
51
+ "style2id": {
52
+ "Neutral": 0,
53
+ "ε…ƒζ°—": 1,
54
+ "呼びかけ": 2,
55
+ "困惑": 3,
56
+ "戦場": 4,
57
+ "ζ„Ÿζƒ…θ±Šγ‹": 5
58
+ }
59
+ },
60
+ "model": {
61
+ "use_spk_conditioned_encoder": true,
62
+ "use_noise_scaled_mas": true,
63
+ "use_mel_posterior_encoder": false,
64
+ "use_duration_discriminator": false,
65
+ "use_wavlm_discriminator": true,
66
+ "inter_channels": 192,
67
+ "hidden_channels": 192,
68
+ "filter_channels": 768,
69
+ "n_heads": 2,
70
+ "n_layers": 6,
71
+ "kernel_size": 3,
72
+ "p_dropout": 0.1,
73
+ "resblock": "1",
74
+ "resblock_kernel_sizes": [
75
+ 3,
76
+ 7,
77
+ 11
78
+ ],
79
+ "resblock_dilation_sizes": [
80
+ [
81
+ 1,
82
+ 3,
83
+ 5
84
+ ],
85
+ [
86
+ 1,
87
+ 3,
88
+ 5
89
+ ],
90
+ [
91
+ 1,
92
+ 3,
93
+ 5
94
+ ]
95
+ ],
96
+ "upsample_rates": [
97
+ 8,
98
+ 8,
99
+ 2,
100
+ 2,
101
+ 2
102
+ ],
103
+ "upsample_initial_channel": 512,
104
+ "upsample_kernel_sizes": [
105
+ 16,
106
+ 16,
107
+ 8,
108
+ 2,
109
+ 2
110
+ ],
111
+ "n_layers_q": 3,
112
+ "use_spectral_norm": false,
113
+ "gin_channels": 512,
114
+ "slm": {
115
+ "model": "./slm/wavlm-base-plus",
116
+ "sr": 16000,
117
+ "hidden": 768,
118
+ "nlayers": 13,
119
+ "initial_channel": 64
120
+ }
121
+ },
122
+ "version": "2.5.0-JP-Extra"
123
+ }
model_assets/rize/rize_e100_s1700.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9cef9207d406d418848694a2887d042139b7563f4b6aaadaf0d517528fffec91
3
+ size 251150980
model_assets/rize/rize_e59_s1000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8640c30e21ec4a932c21db4a037b940f2843b6a0e0d9dfb43893eda12befe13a
3
+ size 251150980
model_assets/rize/style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45ba5f878a2f4507e59355264193c5e658d4a49308b1bd9bd1ca13cf4131b042
3
+ size 6272