Hugging Face's logo

goryden
/
vits2-mn

TensorBoard
Model card Files
xet
Metrics Community
goryden commited on
Commit
945bf0b
·
verified ·
1 Parent(s): 95337f9

Upload folder using huggingface_hub

Browse files
Files changed (11) hide show
  1. DUR_0.pth +3 -0
  2. DUR_1000.pth +3 -0
  3. D_0.pth +3 -0
  4. D_1000.pth +3 -0
  5. G_0.pth +3 -0
  6. G_1000.pth +3 -0
  7. config.json +116 -0
  8. eval/events.out.tfevents.1753861323.d0a2e4808659.339.1 +3 -0
  9. events.out.tfevents.1753861323.d0a2e4808659.339.0 +3 -0
  10. githash +1 -0
  11. train.log +63 -0
DUR_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4a5d00d4732d741f3a1113467c85f7acb4f8c01316f49a1018bde39191f62b2
3
+ size 1695870
DUR_1000.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c896972eb53f69c3e7ef4867d240576de297be1ca267a0e9b9acc2b75ab7021d
3
+ size 1699770
D_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a0232e47fd5b18b51b4a2aaebbccb1cce833ffdd3ec8a09ee2f289404d7247c
3
+ size 561078330
D_1000.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57f18784b5f2c7506b91a4315930d77cfa08a6c68a875b31f5e0d9713eac318d
3
+ size 561093690
G_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:351da17ada47f590dae1b8c80daaddf329551a8e9fb79ba0defb43a30b43d18b
3
+ size 114876142
G_1000.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec2c591580bb7f00b21ae1228e2d59d0107c91fcc595e8b73e21a0c1bd4065b6
3
+ size 114966606
config.json ADDED
@@ -0,0 +1,116 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "log_interval": 200,
4
+ "eval_interval": 1000,
5
+ "seed": 1234,
6
+ "epochs": 1000,
7
+ "learning_rate": 0.0002,
8
+ "betas": [
9
+ 0.8,
10
+ 0.99
11
+ ],
12
+ "eps": 1e-09,
13
+ "batch_size": 32,
14
+ "fp16_run": false,
15
+ "lr_decay": 0.999875,
16
+ "segment_size": 8192,
17
+ "init_lr_ratio": 1,
18
+ "warmup_epochs": 0,
19
+ "c_mel": 45,
20
+ "c_kl": 1.0,
21
+ "fft_sizes": [
22
+ 384,
23
+ 683,
24
+ 171
25
+ ],
26
+ "hop_sizes": [
27
+ 30,
28
+ 60,
29
+ 10
30
+ ],
31
+ "win_lengths": [
32
+ 150,
33
+ 300,
34
+ 60
35
+ ],
36
+ "window": "hann_window"
37
+ },
38
+ "data": {
39
+ "use_mel_posterior_encoder": true,
40
+ "training_files": "filelists/train.txt.cleaned",
41
+ "validation_files": "filelists/val.txt.cleaned",
42
+ "text_cleaners": [
43
+ "mongolian_cleaners"
44
+ ],
45
+ "max_wav_value": 32768.0,
46
+ "sampling_rate": 22050,
47
+ "filter_length": 1024,
48
+ "hop_length": 256,
49
+ "win_length": 1024,
50
+ "n_mel_channels": 80,
51
+ "mel_fmin": 0.0,
52
+ "mel_fmax": null,
53
+ "add_blank": true,
54
+ "n_speakers": 0,
55
+ "cleaned_text": true
56
+ },
57
+ "model": {
58
+ "model_dir": "/kaggle/working/MB-iSTFT-VITS2/logs/models/test",
59
+ "monotonic_align": "sma_v1",
60
+ "use_mel_posterior_encoder": true,
61
+ "use_transformer_flows": true,
62
+ "transformer_flow_type": "pre_conv2",
63
+ "use_spk_conditioned_encoder": false,
64
+ "use_noise_scaled_mas": true,
65
+ "use_duration_discriminator": true,
66
+ "duration_discriminator_type": "dur_disc_2",
67
+ "ms_istft_vits": false,
68
+ "mb_istft_vits": false,
69
+ "istft_vits": true,
70
+ "subbands": false,
71
+ "gen_istft_n_fft": 16,
72
+ "gen_istft_hop_size": 4,
73
+ "inter_channels": 192,
74
+ "hidden_channels": 96,
75
+ "filter_channels": 768,
76
+ "n_heads": 2,
77
+ "n_layers": 3,
78
+ "kernel_size": 3,
79
+ "p_dropout": 0.1,
80
+ "resblock": "1",
81
+ "resblock_kernel_sizes": [
82
+ 3,
83
+ 7,
84
+ 11
85
+ ],
86
+ "resblock_dilation_sizes": [
87
+ [
88
+ 1,
89
+ 3,
90
+ 5
91
+ ],
92
+ [
93
+ 1,
94
+ 3,
95
+ 5
96
+ ],
97
+ [
98
+ 1,
99
+ 3,
100
+ 5
101
+ ]
102
+ ],
103
+ "upsample_rates": [
104
+ 8,
105
+ 8
106
+ ],
107
+ "upsample_initial_channel": 256,
108
+ "upsample_kernel_sizes": [
109
+ 16,
110
+ 16
111
+ ],
112
+ "n_layers_q": 3,
113
+ "use_spectral_norm": false,
114
+ "use_sdp": false
115
+ }
116
+ }
eval/events.out.tfevents.1753861323.d0a2e4808659.339.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d9425c02782395cb9e5624458a877d804f68525cf6d1a89a814a5f833216f4b
3
+ size 1585884
events.out.tfevents.1753861323.d0a2e4808659.339.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5419a18ef0702b30a2e83af9772e782c599dc2670ae86178d08deb85a7484a96
3
+ size 1361797
githash ADDED
@@ -0,0 +1 @@
 
 
1
+ b1a49d9ac2cf6541f28ebf75b112b61afd2a38f1
train.log ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 2025-07-30 07:42:03,742 test INFO {'train': {'log_interval': 200, 'eval_interval': 1000, 'seed': 1234, 'epochs': 1000, 'learning_rate': 0.0002, 'betas': [0.8, 0.99], 'eps': 1e-09, 'batch_size': 32, 'fp16_run': False, 'lr_decay': 0.999875, 'segment_size': 8192, 'init_lr_ratio': 1, 'warmup_epochs': 0, 'c_mel': 45, 'c_kl': 1.0, 'fft_sizes': [384, 683, 171], 'hop_sizes': [30, 60, 10], 'win_lengths': [150, 300, 60], 'window': 'hann_window'}, 'data': {'use_mel_posterior_encoder': True, 'training_files': 'filelists/train.txt.cleaned', 'validation_files': 'filelists/val.txt.cleaned', 'text_cleaners': ['mongolian_cleaners'], 'max_wav_value': 32768.0, 'sampling_rate': 22050, 'filter_length': 1024, 'hop_length': 256, 'win_length': 1024, 'n_mel_channels': 80, 'mel_fmin': 0.0, 'mel_fmax': None, 'add_blank': True, 'n_speakers': 0, 'cleaned_text': True}, 'model': {'model_dir': '/kaggle/working/MB-iSTFT-VITS2/logs/models/test', 'monotonic_align': 'sma_v1', 'use_mel_posterior_encoder': True, 'use_transformer_flows': True, 'transformer_flow_type': 'pre_conv2', 'use_spk_conditioned_encoder': False, 'use_noise_scaled_mas': True, 'use_duration_discriminator': True, 'duration_discriminator_type': 'dur_disc_2', 'ms_istft_vits': False, 'mb_istft_vits': False, 'istft_vits': True, 'subbands': False, 'gen_istft_n_fft': 16, 'gen_istft_hop_size': 4, 'inter_channels': 192, 'hidden_channels': 96, 'filter_channels': 768, 'n_heads': 2, 'n_layers': 3, 'kernel_size': 3, 'p_dropout': 0.1, 'resblock': '1', 'resblock_kernel_sizes': [3, 7, 11], 'resblock_dilation_sizes': [[1, 3, 5], [1, 3, 5], [1, 3, 5]], 'upsample_rates': [8, 8], 'upsample_initial_channel': 256, 'upsample_kernel_sizes': [16, 16], 'n_layers_q': 3, 'use_spectral_norm': False, 'use_sdp': False}, 'model_dir': './logs/models/test'}
2
+ 2025-07-30 07:43:59,191 test INFO Train Epoch: 1 [0%]
3
+ 2025-07-30 07:43:59,192 test INFO [6.0034499168396, 4.460074424743652, 0.44041386246681213, 103.89593505859375, 1.1179826259613037, 400.7793884277344, 0.0, 0, 0.0002]
4
+ 2025-07-30 07:44:11,535 test INFO Saving model and optimizer state at iteration 1 to ./logs/models/test/G_0.pth
5
+ 2025-07-30 07:44:11,899 test INFO Saving model and optimizer state at iteration 1 to ./logs/models/test/D_0.pth
6
+ 2025-07-30 07:44:12,853 test INFO Saving model and optimizer state at iteration 1 to ./logs/models/test/DUR_0.pth
7
+ 2025-07-30 07:47:48,410 test INFO ====> Epoch: 1
8
+ 2025-07-30 07:52:16,097 test INFO ====> Epoch: 2
9
+ 2025-07-30 07:56:26,373 test INFO ====> Epoch: 3
10
+ 2025-07-30 07:59:49,818 test INFO Train Epoch: 4 [85%]
11
+ 2025-07-30 07:59:49,818 test INFO [1.5180813074111938, 3.772383689880371, 7.359023571014404, 39.74175262451172, 0.9084119200706482, 2.220526933670044, 0.0, 200, 0.00019992500937460937]
12
+ 2025-07-30 08:00:21,458 test INFO ====> Epoch: 4
13
+ 2025-07-30 08:04:11,738 test INFO ====> Epoch: 5
14
+ 2025-07-30 08:08:06,495 test INFO ====> Epoch: 6
15
+ 2025-07-30 08:11:53,855 test INFO ====> Epoch: 7
16
+ 2025-07-30 08:14:49,281 test INFO Train Epoch: 8 [69%]
17
+ 2025-07-30 08:14:49,281 test INFO [1.665295958518982, 3.1264448165893555, 7.081151008605957, 38.00682067871094, 0.7591341733932495, 1.3968123197555542, 0.0, 400, 0.00019982506561132978]
18
+ 2025-07-30 08:15:43,705 test INFO ====> Epoch: 8
19
+ 2025-07-30 08:19:29,369 test INFO ====> Epoch: 9
20
+ 2025-07-30 08:23:18,135 test INFO ====> Epoch: 10
21
+ 2025-07-30 08:27:03,509 test INFO ====> Epoch: 11
22
+ 2025-07-30 08:29:38,220 test INFO Train Epoch: 12 [54%]
23
+ 2025-07-30 08:29:38,221 test INFO [2.1282832622528076, 3.220754623413086, 5.894423484802246, 31.136409759521484, 0.674907922744751, 1.5520458221435547, 0.0, 600, 0.00019972517181056292]
24
+ 2025-07-30 08:30:50,375 test INFO ====> Epoch: 12
25
+ 2025-07-30 08:34:35,957 test INFO ====> Epoch: 13
26
+ 2025-07-30 08:38:20,543 test INFO ====> Epoch: 14
27
+ 2025-07-30 08:42:04,497 test INFO ====> Epoch: 15
28
+ 2025-07-30 08:44:14,277 test INFO Train Epoch: 16 [38%]
29
+ 2025-07-30 08:44:14,277 test INFO [2.2449469566345215, 2.702267646789551, 4.473440170288086, 28.876434326171875, 0.7084023952484131, 1.6447468996047974, 0.0, 800, 0.00019962532794733217]
30
+ 2025-07-30 08:45:48,232 test INFO ====> Epoch: 16
31
+ 2025-07-30 08:49:32,848 test INFO ====> Epoch: 17
32
+ 2025-07-30 08:53:18,006 test INFO ====> Epoch: 18
33
+ 2025-07-30 08:57:02,739 test INFO ====> Epoch: 19
34
+ 2025-07-30 08:58:52,173 test INFO Train Epoch: 20 [23%]
35
+ 2025-07-30 08:58:52,173 test INFO [2.0880696773529053, 2.853564977645874, 5.468815326690674, 27.427783966064453, 0.6990370750427246, 1.5836814641952515, 0.0, 1000, 0.00019952553399667344]
36
+ 2025-07-30 08:59:03,185 test INFO Saving model and optimizer state at iteration 20 to ./logs/models/test/G_1000.pth
37
+ 2025-07-30 08:59:03,478 test INFO Saving model and optimizer state at iteration 20 to ./logs/models/test/D_1000.pth
38
+ 2025-07-30 08:59:04,403 test INFO Saving model and optimizer state at iteration 20 to ./logs/models/test/DUR_1000.pth
39
+ 2025-07-30 09:01:00,895 test INFO ====> Epoch: 20
40
+ 2025-07-30 09:04:47,141 test INFO ====> Epoch: 21
41
+ 2025-07-30 09:08:38,819 test INFO ====> Epoch: 22
42
+ 2025-07-30 09:12:26,754 test INFO ====> Epoch: 23
43
+ 2025-07-30 09:13:52,984 test INFO Train Epoch: 24 [8%]
44
+ 2025-07-30 09:13:52,984 test INFO [2.16131854057312, 3.40720796585083, 5.75662899017334, 28.300180435180664, 0.6504846811294556, 1.7258610725402832, 0.0, 1200, 0.00019942578993363514]
45
+ 2025-07-30 09:16:16,551 test INFO ====> Epoch: 24
46
+ 2025-07-30 09:20:05,126 test INFO ====> Epoch: 25
47
+ 2025-07-30 09:23:51,113 test INFO ====> Epoch: 26
48
+ 2025-07-30 09:27:21,732 test INFO Train Epoch: 27 [92%]
49
+ 2025-07-30 09:27:21,732 test INFO [2.3914968967437744, 3.457401990890503, 4.960050106048584, 25.98041343688965, 0.6626256704330444, 1.799747347831726, 0.0, 1400, 0.00019935101461010442]
50
+ 2025-07-30 09:27:40,225 test INFO ====> Epoch: 27
51
+ 2025-07-30 09:31:25,852 test INFO ====> Epoch: 28
52
+ 2025-07-30 09:35:14,860 test INFO ====> Epoch: 29
53
+ 2025-07-30 09:39:01,049 test INFO ====> Epoch: 30
54
+ 2025-07-30 09:42:10,772 test INFO Train Epoch: 31 [77%]
55
+ 2025-07-30 09:42:10,773 test INFO [2.2855846881866455, 2.450014352798462, 4.642941474914551, 25.688888549804688, 0.7220066785812378, 1.6248011589050293, 0.0, 1600, 0.00019925135779039958]
56
+ 2025-07-30 09:42:50,045 test INFO ====> Epoch: 31
57
+ 2025-07-30 09:46:37,840 test INFO ====> Epoch: 32
58
+ 2025-07-30 09:50:26,613 test INFO ====> Epoch: 33
59
+ 2025-07-30 09:54:15,167 test INFO ====> Epoch: 34
60
+ 2025-07-30 09:57:01,649 test INFO Train Epoch: 35 [62%]
61
+ 2025-07-30 09:57:01,650 test INFO [2.525804281234741, 2.3113038539886475, 3.826836585998535, 24.56623077392578, 0.9666768312454224, 1.6983153820037842, 0.0, 1800, 0.00019915175078976256]
62
+ 2025-07-30 09:58:04,870 test INFO ====> Epoch: 35
63
+ 2025-07-30 10:01:50,139 test INFO ====> Epoch: 36