Hugging Face's logo

rufflet17
/
voice_models

Model card Files
xet
Community
rufflet17 commited on
Commit
c9a2b2b
·
verified ·
1 Parent(s): fa72242

Update CO/SZ/config.json

Browse files
Files changed (1) hide show
  1. CO/SZ/config.json +121 -121
CO/SZ/config.json CHANGED
@@ -1,122 +1,122 @@
1
- {
2
- "model_name": "SZ",
3
- "train": {
4
- "log_interval": 200,
5
- "eval_interval": 1000,
6
- "seed": 42,
7
- "epochs": 1000,
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/SZ/train.list",
35
- "validation_files": "Data/SZ/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
- "SZ": 0
49
- },
50
- "num_styles": 5,
51
- "style2id": {
52
- "Neutral": 0,
53
- "ちょっと高め": 1,
54
- "普通１": 2,
55
- "普通２": 3,
56
- "普通３": 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": "2.6.1-JP-Extra"
122
  }
 
1
+ {
2
+ "model_name": "SZ",
3
+ "train": {
4
+ "log_interval": 200,
5
+ "eval_interval": 1000,
6
+ "seed": 42,
7
+ "epochs": 1000,
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/SZ/train.list",
35
+ "validation_files": "Data/SZ/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
+ "SZ": 0
49
+ },
50
+ "num_styles": 5,
51
+ "style2id": {
52
+ "Neutral": 0,
53
+ "1": 1,
54
+ "2": 2,
55
+ "3": 3,
56
+ "4": 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": "2.6.1-JP-Extra"
122
  }