v3ucn commited on
Commit
cd327b9
1 Parent(s): f9222b5

commit from liuyue

Browse files
Files changed (2) hide show
  1. G_200.pth +3 -0
  2. config.json +111 -0
G_200.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd7c5df76386134683ad522a055a56d0643242868b47ecd2d9264754b3106bc0
3
+ size 926450974
config.json ADDED
@@ -0,0 +1,111 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "log_interval": 200,
4
+ "eval_interval": 200,
5
+ "seed": 42,
6
+ "epochs": 1000,
7
+ "learning_rate": 0.0002,
8
+ "betas": [
9
+ 0.8,
10
+ 0.99
11
+ ],
12
+ "eps": 1e-09,
13
+ "batch_size": 6,
14
+ "bf16_run": false,
15
+ "fp16_run": false,
16
+ "lr_decay": 0.99996,
17
+ "segment_size": 16384,
18
+ "init_lr_ratio": 1,
19
+ "warmup_epochs": 0,
20
+ "c_mel": 45,
21
+ "c_kl": 1.0,
22
+ "c_commit": 100,
23
+ "skip_optimizer": true,
24
+ "freeze_ZH_bert": false,
25
+ "freeze_JP_bert": false,
26
+ "freeze_EN_bert": false,
27
+ "freeze_emo": false
28
+ },
29
+ "data": {
30
+ "training_files": "data/Wizard3/train.list",
31
+ "validation_files": "data/Wizard3/val.list",
32
+ "max_wav_value": 32768.0,
33
+ "sampling_rate": 44100,
34
+ "filter_length": 2048,
35
+ "hop_length": 512,
36
+ "win_length": 2048,
37
+ "n_mel_channels": 128,
38
+ "mel_fmin": 0.0,
39
+ "mel_fmax": null,
40
+ "add_blank": true,
41
+ "n_speakers": 2,
42
+ "cleaned_text": true,
43
+ "spk2id": {
44
+ "Keira": 0,
45
+ "Yennefer": 1
46
+ }
47
+ },
48
+ "model": {
49
+ "use_spk_conditioned_encoder": true,
50
+ "use_noise_scaled_mas": true,
51
+ "use_mel_posterior_encoder": false,
52
+ "use_duration_discriminator": false,
53
+ "use_wavlm_discriminator": true,
54
+ "inter_channels": 192,
55
+ "hidden_channels": 192,
56
+ "filter_channels": 768,
57
+ "n_heads": 2,
58
+ "n_layers": 6,
59
+ "kernel_size": 3,
60
+ "p_dropout": 0.1,
61
+ "resblock": "1",
62
+ "resblock_kernel_sizes": [
63
+ 3,
64
+ 7,
65
+ 11
66
+ ],
67
+ "resblock_dilation_sizes": [
68
+ [
69
+ 1,
70
+ 3,
71
+ 5
72
+ ],
73
+ [
74
+ 1,
75
+ 3,
76
+ 5
77
+ ],
78
+ [
79
+ 1,
80
+ 3,
81
+ 5
82
+ ]
83
+ ],
84
+ "upsample_rates": [
85
+ 8,
86
+ 8,
87
+ 2,
88
+ 2,
89
+ 2
90
+ ],
91
+ "upsample_initial_channel": 512,
92
+ "upsample_kernel_sizes": [
93
+ 16,
94
+ 16,
95
+ 8,
96
+ 2,
97
+ 2
98
+ ],
99
+ "n_layers_q": 3,
100
+ "use_spectral_norm": false,
101
+ "gin_channels": 256,
102
+ "slm": {
103
+ "model": "./slm/wavlm-base-plus",
104
+ "sr": 16000,
105
+ "hidden": 768,
106
+ "nlayers": 13,
107
+ "initial_channel": 64
108
+ }
109
+ },
110
+ "version": "2.3"
111
+ }