Hugging Face's logo

dog
/
kanye

Audio-to-Audio
PyTorch
TensorBoard
English
so-vits-svc
voice-cloning
Model card Files Metrics Community
3
dog commited on
Commit
eded570
•
1 Parent(s): 0463d9f

🍻 cheers

Browse files
Files changed (7) hide show
  1. D_0.pth +3 -0
  2. D_84.pth +3 -0
  3. G_0.pth +3 -0
  4. G_84.pth +3 -0
  5. config.json +102 -0
  6. lightning_logs/version_0/events.out.tfevents.1681978824.a3d43af43c98.11299.0 +3 -0
  7. lightning_logs/version_0/hparams.yaml +87 -0
D_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:635be5c3409aaf3eec4135a1f5a771595683f3a6461ffc5bdea43441e50269a9
3
+ size 187018591
D_84.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:143f76b07c03d4de0456e7061c579da02a98aa45b3e8987b267c165a33eaad19
3
+ size 561098249
G_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20a327c54e5731bed377bd38404bc32ab98e66a1b2777b0af4cc034d4d6914b0
3
+ size 180628517
G_84.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6073adf19c5f3ebd1375911754cd666bb7d7b12a870d4fc835f253ea5fb3ce44
3
+ size 542789469
config.json ADDED
@@ -0,0 +1,102 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "log_interval": 200,
4
+ "eval_interval": 500,
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": 32,
14
+ "fp16_run": false,
15
+ "bf16_run": false,
16
+ "lr_decay": 0.999875,
17
+ "segment_size": 10240,
18
+ "init_lr_ratio": 1,
19
+ "warmup_epochs": 0,
20
+ "c_mel": 45,
21
+ "c_kl": 1.0,
22
+ "use_sr": true,
23
+ "max_speclen": 512,
24
+ "port": "8001",
25
+ "keep_ckpts": 3,
26
+ "num_workers": 12,
27
+ "log_version": 0,
28
+ "ckpt_name_by_step": false,
29
+ "accumulate_grad_batches": 1,
30
+ "persistent_workers": true,
31
+ "push_to_hub": true,
32
+ "repo_id": "dog/kanye",
33
+ "private": true
34
+ },
35
+ "data": {
36
+ "training_files": "filelists/44k/train.txt",
37
+ "validation_files": "filelists/44k/val.txt",
38
+ "max_wav_value": 32768.0,
39
+ "sampling_rate": 44100,
40
+ "filter_length": 2048,
41
+ "hop_length": 512,
42
+ "win_length": 2048,
43
+ "n_mel_channels": 80,
44
+ "mel_fmin": 0.0,
45
+ "mel_fmax": 22050
46
+ },
47
+ "model": {
48
+ "inter_channels": 192,
49
+ "hidden_channels": 192,
50
+ "filter_channels": 768,
51
+ "n_heads": 2,
52
+ "n_layers": 6,
53
+ "kernel_size": 3,
54
+ "p_dropout": 0.1,
55
+ "resblock": "1",
56
+ "resblock_kernel_sizes": [
57
+ 3,
58
+ 7,
59
+ 11
60
+ ],
61
+ "resblock_dilation_sizes": [
62
+ [
63
+ 1,
64
+ 3,
65
+ 5
66
+ ],
67
+ [
68
+ 1,
69
+ 3,
70
+ 5
71
+ ],
72
+ [
73
+ 1,
74
+ 3,
75
+ 5
76
+ ]
77
+ ],
78
+ "upsample_rates": [
79
+ 8,
80
+ 8,
81
+ 2,
82
+ 2,
83
+ 2
84
+ ],
85
+ "upsample_initial_channel": 512,
86
+ "upsample_kernel_sizes": [
87
+ 16,
88
+ 16,
89
+ 4,
90
+ 4,
91
+ 4
92
+ ],
93
+ "n_layers_q": 3,
94
+ "use_spectral_norm": false,
95
+ "gin_channels": 256,
96
+ "ssl_dim": 256,
97
+ "n_speakers": 200
98
+ },
99
+ "spk": {
100
+ "kanye": 0
101
+ }
102
+ }
lightning_logs/version_0/events.out.tfevents.1681978824.a3d43af43c98.11299.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b473aa43ca2d2be2dbb2243b58c9080cbd736787cab19f175f2c20438322082
3
+ size 3012116
lightning_logs/version_0/hparams.yaml ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ data: !!python/object:so_vits_svc_fork.hparams.HParams
2
+ filter_length: 2048
3
+ hop_length: 512
4
+ max_wav_value: 32768.0
5
+ mel_fmax: 22050
6
+ mel_fmin: 0.0
7
+ n_mel_channels: 80
8
+ sampling_rate: 44100
9
+ training_files: filelists/44k/train.txt
10
+ validation_files: filelists/44k/val.txt
11
+ win_length: 2048
12
+ model: !!python/object:so_vits_svc_fork.hparams.HParams
13
+ filter_channels: 768
14
+ gin_channels: 256
15
+ hidden_channels: 192
16
+ inter_channels: 192
17
+ kernel_size: 3
18
+ n_heads: 2
19
+ n_layers: 6
20
+ n_layers_q: 3
21
+ n_speakers: 200
22
+ p_dropout: 0.1
23
+ resblock: '1'
24
+ resblock_dilation_sizes:
25
+ - - 1
26
+ - 3
27
+ - 5
28
+ - - 1
29
+ - 3
30
+ - 5
31
+ - - 1
32
+ - 3
33
+ - 5
34
+ resblock_kernel_sizes:
35
+ - 3
36
+ - 7
37
+ - 11
38
+ ssl_dim: 256
39
+ upsample_initial_channel: 512
40
+ upsample_kernel_sizes:
41
+ - 16
42
+ - 16
43
+ - 4
44
+ - 4
45
+ - 4
46
+ upsample_rates:
47
+ - 8
48
+ - 8
49
+ - 2
50
+ - 2
51
+ - 2
52
+ use_spectral_norm: false
53
+ model_dir: logs/44k
54
+ reset_optimizer: false
55
+ spk: !!python/object:so_vits_svc_fork.hparams.HParams
56
+ kanye: 0
57
+ train: !!python/object:so_vits_svc_fork.hparams.HParams
58
+ accumulate_grad_batches: 1
59
+ batch_size: 32
60
+ betas:
61
+ - 0.8
62
+ - 0.99
63
+ bf16_run: false
64
+ c_kl: 1.0
65
+ c_mel: 45
66
+ ckpt_name_by_step: false
67
+ epochs: 10000
68
+ eps: 1.0e-09
69
+ eval_interval: 500
70
+ fp16_run: false
71
+ init_lr_ratio: 1
72
+ keep_ckpts: 3
73
+ learning_rate: 0.0001
74
+ log_interval: 200
75
+ log_version: 0
76
+ lr_decay: 0.999875
77
+ max_speclen: 512
78
+ num_workers: 12
79
+ persistent_workers: true
80
+ port: '8001'
81
+ private: true
82
+ push_to_hub: true
83
+ repo_id: dog/kanye
84
+ seed: 1234
85
+ segment_size: 10240
86
+ use_sr: true
87
+ warmup_epochs: 0