Hugging Face's logo

Spaces:
mzltest
/
test-vits3
Build error

App Files Community
mzltest commited on
Commit
d2805cb
·
verified ·
1 Parent(s): 3d4893c

Update config.json

Browse files
Files changed (1) hide show
  1. config.json +88 -90
config.json CHANGED
@@ -1,92 +1,90 @@
1
  {
2
- "train": {
3
- "log_interval": 200,
4
- "eval_interval": 1000,
5
- "seed": 1234,
6
- "epochs": 10000,
7
- "learning_rate": 0.0001,
8
- "betas": [
9
- 0.8,
10
- 0.99
11
- ],
12
- "eps": 1e-9,
13
- "batch_size": 12,
14
- "fp16_run": false,
15
- "lr_decay": 0.999875,
16
- "segment_size": 17920,
17
- "init_lr_ratio": 1,
18
- "warmup_epochs": 0,
19
- "c_mel": 45,
20
- "c_kl": 1,
21
- "use_sr": true,
22
- "max_speclen": 384,
23
- "port": "8001"
24
- },
25
- "data": {
26
- "training_files": "filelists/train.txt",
27
- "validation_files": "filelists/val.txt",
28
- "max_wav_value": 32768,
29
- "sampling_rate": 32000,
30
- "filter_length": 1280,
31
- "hop_length": 320,
32
- "win_length": 1280,
33
- "n_mel_channels": 80,
34
- "mel_fmin": 0,
35
- "mel_fmax": null
36
- },
37
- "model": {
38
- "inter_channels": 192,
39
- "hidden_channels": 192,
40
- "filter_channels": 768,
41
- "n_heads": 2,
42
- "n_layers": 6,
43
- "kernel_size": 3,
44
- "p_dropout": 0.1,
45
- "resblock": "1",
46
- "resblock_kernel_sizes": [
47
- 3,
48
- 7,
49
- 11
50
- ],
51
- "resblock_dilation_sizes": [
52
- [
53
- 1,
54
- 3,
55
- 5
56
- ],
57
- [
58
- 1,
59
- 3,
60
- 5
61
- ],
62
- [
63
- 1,
64
- 3,
65
- 5
66
- ]
67
- ],
68
- "upsample_rates": [
69
- 10,
70
- 8,
71
- 2,
72
- 2
73
- ],
74
- "upsample_initial_channel": 512,
75
- "upsample_kernel_sizes": [
76
- 16,
77
- 16,
78
- 4,
79
- 4
80
- ],
81
- "n_layers_q": 3,
82
- "use_spectral_norm": false,
83
- "gin_channels": 256,
84
- "ssl_dim": 256,
85
- "n_speakers": 6
86
- },
87
- "spk": {
88
- "haruka": 0,
89
- "maya": 1,
90
- "koyomi": 2
91
- }
92
  }
 
1
  {
2
+ "train": {
3
+ "log_interval": 200,
4
+ "eval_interval": 1000,
5
+ "seed": 1234,
6
+ "epochs": 10000,
7
+ "learning_rate": 0.0001,
8
+ "betas": [
9
+ 0.8,
10
+ 0.99
11
+ ],
12
+ "eps": 1e-09,
13
+ "batch_size": 12,
14
+ "fp16_run": false,
15
+ "lr_decay": 0.999875,
16
+ "segment_size": 17920,
17
+ "init_lr_ratio": 1,
18
+ "warmup_epochs": 0,
19
+ "c_mel": 45,
20
+ "c_kl": 1.0,
21
+ "use_sr": true,
22
+ "max_speclen": 384,
23
+ "port": "8001"
24
+ },
25
+ "data": {
26
+ "training_files": "filelists/train.txt",
27
+ "validation_files": "filelists/val.txt",
28
+ "max_wav_value": 32768.0,
29
+ "sampling_rate": 32000,
30
+ "filter_length": 1280,
31
+ "hop_length": 320,
32
+ "win_length": 1280,
33
+ "n_mel_channels": 80,
34
+ "mel_fmin": 0.0,
35
+ "mel_fmax": null
36
+ },
37
+ "model": {
38
+ "inter_channels": 192,
39
+ "hidden_channels": 192,
40
+ "filter_channels": 768,
41
+ "n_heads": 2,
42
+ "n_layers": 6,
43
+ "kernel_size": 3,
44
+ "p_dropout": 0.1,
45
+ "resblock": "1",
46
+ "resblock_kernel_sizes": [
47
+ 3,
48
+ 7,
49
+ 11
50
+ ],
51
+ "resblock_dilation_sizes": [
52
+ [
53
+ 1,
54
+ 3,
55
+ 5
56
+ ],
57
+ [
58
+ 1,
59
+ 3,
60
+ 5
61
+ ],
62
+ [
63
+ 1,
64
+ 3,
65
+ 5
66
+ ]
67
+ ],
68
+ "upsample_rates": [
69
+ 10,
70
+ 8,
71
+ 2,
72
+ 2
73
+ ],
74
+ "upsample_initial_channel": 512,
75
+ "upsample_kernel_sizes": [
76
+ 16,
77
+ 16,
78
+ 4,
79
+ 4
80
+ ],
81
+ "n_layers_q": 3,
82
+ "use_spectral_norm": false,
83
+ "gin_channels": 256,
84
+ "ssl_dim": 256,
85
+ "n_speakers": 2
86
+ },
87
+ "spk": {
88
+ "chiya": 0
89
+ }
 
 
90
  }