Hugging Face's logo

Yukiyoke-Lab
/
Tsukuyomi-chan

Text-to-Speech
ONNX
Japanese
Model card Files
xet
 Community
2
akikukeo commited on
Commit
961aff1
·
1 Parent(s): 1192695

add models

Browse files
Files changed (10) hide show
  1. Tsukuyomi_e116_s6000.safetensors +3 -0
  2. Tsukuyomi_e20_s1000.safetensors +3 -0
  3. Tsukuyomi_e39_s2000.safetensors +3 -0
  4. Tsukuyomi_e58_s3000.onnx +3 -0
  5. Tsukuyomi_e58_s3000.safetensors +3 -0
  6. Tsukuyomi_e78_s4000.safetensors +3 -0
  7. Tsukuyomi_e97_s5000.onnx +3 -0
  8. Tsukuyomi_e97_s5000.safetensors +3 -0
  9. config.json +118 -0
  10. style_vectors.npy +3 -0
Tsukuyomi_e116_s6000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01a49bbfa5ddf787fc0e57592a29c4abbd360d510b1dce3891fd0384dd2b4324
3
+ size 251150980
Tsukuyomi_e20_s1000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b12e6c01906dc1579c901b32a6016a2d2529acd6619f18872e409595f9fb824
3
+ size 251150980
Tsukuyomi_e39_s2000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01e8302a874ece6b489c95bf387581e35f77277b5f9818a737635ac3fb713785
3
+ size 251150980
Tsukuyomi_e58_s3000.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a0cc6b7f71182c22f3e02e7dabf189a48443629397690b54619dc0b56b3f20ba
3
+ size 249417245
Tsukuyomi_e58_s3000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0b96cce8c95799f3b26e52409d4356b46e8fcdc2a554e18d0ba3930f335a0b5
3
+ size 251150980
Tsukuyomi_e78_s4000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1e92e43507e308970d41b755654e518756ba78287c58121ba242b0fd7082e67c
3
+ size 251150980
Tsukuyomi_e97_s5000.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f94175e889dbd23bfd46a896f50d9227ce6ea9f364bb0484b411251691d7974
3
+ size 249417245
Tsukuyomi_e97_s5000.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50c71eb854ae52903039eee101cac734d7c6a161036e538175d2472e22e247f8
3
+ size 251150980
config.json ADDED
@@ -0,0 +1,118 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "Tsukuyomi-tyan",
3
+ "train": {
4
+ "log_interval": 50,
5
+ "eval_interval": 1000,
6
+ "seed": 42,
7
+ "epochs": 300,
8
+ "learning_rate": 0.0001,
9
+ "betas": [
10
+ 0.8,
11
+ 0.99
12
+ ],
13
+ "eps": 1e-09,
14
+ "batch_size": 2,
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": "Data\\Tsukuyomi-tyan\\train.list",
35
+ "validation_files": "Data\\Tsukuyomi-tyan\\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
+ "Tsukuyomi": 0
49
+ },
50
+ "num_styles": 1,
51
+ "style2id": {
52
+ "Neutral": 0
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": false,
60
+ "use_wavlm_discriminator": true,
61
+ "inter_channels": 192,
62
+ "hidden_channels": 192,
63
+ "filter_channels": 768,
64
+ "n_heads": 2,
65
+ "n_layers": 6,
66
+ "kernel_size": 3,
67
+ "p_dropout": 0.1,
68
+ "resblock": "1",
69
+ "resblock_kernel_sizes": [
70
+ 3,
71
+ 7,
72
+ 11
73
+ ],
74
+ "resblock_dilation_sizes": [
75
+ [
76
+ 1,
77
+ 3,
78
+ 5
79
+ ],
80
+ [
81
+ 1,
82
+ 3,
83
+ 5
84
+ ],
85
+ [
86
+ 1,
87
+ 3,
88
+ 5
89
+ ]
90
+ ],
91
+ "upsample_rates": [
92
+ 8,
93
+ 8,
94
+ 2,
95
+ 2,
96
+ 2
97
+ ],
98
+ "upsample_initial_channel": 512,
99
+ "upsample_kernel_sizes": [
100
+ 16,
101
+ 16,
102
+ 8,
103
+ 2,
104
+ 2
105
+ ],
106
+ "n_layers_q": 3,
107
+ "use_spectral_norm": false,
108
+ "gin_channels": 512,
109
+ "slm": {
110
+ "model": "./slm/wavlm-base-plus",
111
+ "sr": 16000,
112
+ "hidden": 768,
113
+ "nlayers": 13,
114
+ "initial_channel": 64
115
+ }
116
+ },
117
+ "version": "2.7.0-JP-Extra"
118
+ }
style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:23c70c534f1ab013e92ed892612e71e5c52479fea00f20796a8fc2dd6a7371aa
3
+ size 1152