cockolo terada commited on
Commit
e1f8b04
·
verified ·
1 Parent(s): c41bd34

Upload CO/RI/config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. CO/RI/config.json +124 -0
CO/RI/config.json ADDED
@@ -0,0 +1,124 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "RI",
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": 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/RI/train.list",
35
+ "validation_files": "Data/RI/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
+ "RI": 0
49
+ },
50
+ "num_styles": 7,
51
+ "style2id": {
52
+ "Neutral": 0,
53
+ "1": 1,
54
+ "2": 2,
55
+ "3": 3,
56
+ "4": 4,
57
+ "5": 5,
58
+ "6": 6
59
+ }
60
+ },
61
+ "model": {
62
+ "use_spk_conditioned_encoder": true,
63
+ "use_noise_scaled_mas": true,
64
+ "use_mel_posterior_encoder": false,
65
+ "use_duration_discriminator": false,
66
+ "use_wavlm_discriminator": true,
67
+ "inter_channels": 192,
68
+ "hidden_channels": 192,
69
+ "filter_channels": 768,
70
+ "n_heads": 2,
71
+ "n_layers": 6,
72
+ "kernel_size": 3,
73
+ "p_dropout": 0.1,
74
+ "resblock": "1",
75
+ "resblock_kernel_sizes": [
76
+ 3,
77
+ 7,
78
+ 11
79
+ ],
80
+ "resblock_dilation_sizes": [
81
+ [
82
+ 1,
83
+ 3,
84
+ 5
85
+ ],
86
+ [
87
+ 1,
88
+ 3,
89
+ 5
90
+ ],
91
+ [
92
+ 1,
93
+ 3,
94
+ 5
95
+ ]
96
+ ],
97
+ "upsample_rates": [
98
+ 8,
99
+ 8,
100
+ 2,
101
+ 2,
102
+ 2
103
+ ],
104
+ "upsample_initial_channel": 512,
105
+ "upsample_kernel_sizes": [
106
+ 16,
107
+ 16,
108
+ 8,
109
+ 2,
110
+ 2
111
+ ],
112
+ "n_layers_q": 3,
113
+ "use_spectral_norm": false,
114
+ "gin_channels": 512,
115
+ "slm": {
116
+ "model": "./slm/wavlm-base-plus",
117
+ "sr": 16000,
118
+ "hidden": 768,
119
+ "nlayers": 13,
120
+ "initial_channel": 64
121
+ }
122
+ },
123
+ "version": "2.6.1-JP-Extra"
124
+ }