Kit-Lemonfoot commited on
Commit
587b3fe
1 Parent(s): 28365bd

Upload 3 files

Browse files
model_assets/SBV2_HoloIDFlu/SBV2_HoloIDFlu.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5cce80bcbbe92aeb1a1006b782812f77e153e3cf30c1e5029bedd7f58b10a8f
3
+ size 198769212
model_assets/SBV2_HoloIDFlu/config.json ADDED
@@ -0,0 +1,110 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "SBV2_HoloIDFlu",
3
+ "train": {
4
+ "log_interval": 200,
5
+ "eval_interval": 1000,
6
+ "seed": 42,
7
+ "epochs": 200,
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_HoloIDFlu/train.list",
30
+ "validation_files": "Data/SBV2_HoloIDFlu/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": 2,
41
+ "cleaned_text": true,
42
+ "num_styles": 5,
43
+ "style2id": {
44
+ "Neutral": 0,
45
+ "ZetaSoft": 1,
46
+ "Zeta": 2,
47
+ "ZetaLoud": 3,
48
+ "Ollie": 4
49
+ },
50
+ "spk2id": {
51
+ "KureijiOllie": 0,
52
+ "VestiaZeta": 1
53
+ }
54
+ },
55
+ "model": {
56
+ "use_spk_conditioned_encoder": true,
57
+ "use_noise_scaled_mas": true,
58
+ "use_mel_posterior_encoder": false,
59
+ "use_duration_discriminator": true,
60
+ "inter_channels": 192,
61
+ "hidden_channels": 192,
62
+ "filter_channels": 768,
63
+ "n_heads": 2,
64
+ "n_layers": 6,
65
+ "kernel_size": 3,
66
+ "p_dropout": 0.1,
67
+ "resblock": "1",
68
+ "resblock_kernel_sizes": [
69
+ 3,
70
+ 7,
71
+ 11
72
+ ],
73
+ "resblock_dilation_sizes": [
74
+ [
75
+ 1,
76
+ 3,
77
+ 5
78
+ ],
79
+ [
80
+ 1,
81
+ 3,
82
+ 5
83
+ ],
84
+ [
85
+ 1,
86
+ 3,
87
+ 5
88
+ ]
89
+ ],
90
+ "upsample_rates": [
91
+ 8,
92
+ 8,
93
+ 2,
94
+ 2,
95
+ 2
96
+ ],
97
+ "upsample_initial_channel": 512,
98
+ "upsample_kernel_sizes": [
99
+ 16,
100
+ 16,
101
+ 8,
102
+ 2,
103
+ 2
104
+ ],
105
+ "n_layers_q": 3,
106
+ "use_spectral_norm": false,
107
+ "gin_channels": 256
108
+ },
109
+ "version": "1.3"
110
+ }
model_assets/SBV2_HoloIDFlu/style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69c402a763f3015dcb971c310e44d21c1774fe0100855f675fb9f6cc58e19caf
3
+ size 5248