Kit-Lemonfoot commited on
Commit
b11a0cd
·
verified ·
1 Parent(s): 0569b2c

Upload 3 files

Browse files
model_assets/SBV2_HoloHi/SBV2_HoloHi.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1d24e55d5950e671db59b151636503206722d362cf8fbb18fb559a69a95d139
3
+ size 198774332
model_assets/SBV2_HoloHi/config.json ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "SBV2_HoloHi",
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": true,
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
+ },
28
+ "data": {
29
+ "training_files": "Data/SBV2_HoloHi/train.list",
30
+ "validation_files": "Data/SBV2_HoloHi/val.list",
31
+ "max_wav_value": 32768.0,
32
+ "sampling_rate": 44100,
33
+ "filter_length": 2048,
34
+ "hop_length": 512,
35
+ "win_length": 2048,
36
+ "n_mel_channels": 128,
37
+ "mel_fmin": 0.0,
38
+ "mel_fmax": null,
39
+ "add_blank": true,
40
+ "n_speakers": 7,
41
+ "cleaned_text": true,
42
+ "num_styles": 10,
43
+ "style2id": {
44
+ "Neutral": 0,
45
+ "Fauna": 1,
46
+ "Amelia": 2,
47
+ "MumeiLaugh": 3,
48
+ "Shiori": 4,
49
+ "IRyS": 5,
50
+ "Ina": 6,
51
+ "Gura": 7,
52
+ "Mumei": 8,
53
+ "ShioriLaugh": 9
54
+ },
55
+ "spk2id": {
56
+ "NinomaeInanis": 0,
57
+ "GawrGura": 1,
58
+ "AmeliaWatson": 2,
59
+ "IRyS": 3,
60
+ "CeresFauna": 4,
61
+ "NanashiMumei": 5,
62
+ "ShioriNovella": 6
63
+ }
64
+ },
65
+ "model": {
66
+ "use_spk_conditioned_encoder": true,
67
+ "use_noise_scaled_mas": true,
68
+ "use_mel_posterior_encoder": false,
69
+ "use_duration_discriminator": true,
70
+ "inter_channels": 192,
71
+ "hidden_channels": 192,
72
+ "filter_channels": 768,
73
+ "n_heads": 2,
74
+ "n_layers": 6,
75
+ "kernel_size": 3,
76
+ "p_dropout": 0.1,
77
+ "resblock": "1",
78
+ "resblock_kernel_sizes": [
79
+ 3,
80
+ 7,
81
+ 11
82
+ ],
83
+ "resblock_dilation_sizes": [
84
+ [
85
+ 1,
86
+ 3,
87
+ 5
88
+ ],
89
+ [
90
+ 1,
91
+ 3,
92
+ 5
93
+ ],
94
+ [
95
+ 1,
96
+ 3,
97
+ 5
98
+ ]
99
+ ],
100
+ "upsample_rates": [
101
+ 8,
102
+ 8,
103
+ 2,
104
+ 2,
105
+ 2
106
+ ],
107
+ "upsample_initial_channel": 512,
108
+ "upsample_kernel_sizes": [
109
+ 16,
110
+ 16,
111
+ 8,
112
+ 2,
113
+ 2
114
+ ],
115
+ "n_layers_q": 3,
116
+ "use_spectral_norm": false,
117
+ "gin_channels": 256
118
+ },
119
+ "version": "1.3"
120
+ }
model_assets/SBV2_HoloHi/style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:209866d5cb3a0fd89c152f4687da9c5c1c35f0902b2043f529411b001cfd0d22
3
+ size 10368