clean unused model
Browse files- Data/Azuma/config.json +0 -108
- Data/Azuma/models/G_17900.pth +0 -3
- Data/Azuma/models/G_18200.pth +0 -3
- Data/Azusa/config.json +0 -108
- Data/Azusa/models/G_11300.pth +0 -3
- Data/Carol/config.json +0 -108
- Data/Carol/models/G_12800.pth +0 -3
- Data/Diana/config.json +0 -108
- Data/Diana/models/G_7800.pth +0 -3
- Data/Eileen/config.json +0 -108
- Data/Eileen/models/G_10700.pth +0 -3
- Data/Eileen/models/G_10800.pth +0 -3
- Data/Messi/config.json +0 -108
- Data/Messi/models/G_1800.pth +0 -3
- Data/Nana7mi/config.json +0 -108
- Data/Nana7mi/models/G_10700.pth +0 -3
- Data/Taffy/config.json +0 -108
- Data/Taffy/models/G_11100.pth +0 -3
- Data/XingTong/config.json +0 -108
- Data/XingTong/models/G_11600.pth +0 -3
- Data/badXT/config.json +0 -108
- Data/badXT/models/G_3250.pth +0 -3
- Data/dingzhen/config.json +0 -108
- Data/dingzhen/models/G_2650.pth +0 -3
- Data/nine/config.json +0 -108
- Data/nine/models/G_10700.pth +0 -3
- Data/otto/config.json +0 -108
- Data/otto/models/G_5400.pth +0 -3
- Data/xuan/config.json +0 -108
- Data/xuan/models/G_7200.pth +0 -3
Data/Azuma/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 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": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/Azuma/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/Azuma/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"东雪莲": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/Azuma/models/G_17900.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:bece1cd9129451f15a1b143a7dfaad0ea4d7c5718725aa48825cbf89f4b31983
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/Azuma/models/G_18200.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:fe75c0301cbf361898fc7c9aff122dd411df647b3fc8a8a53c4846f07d5f7a43
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/Azusa/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 5 |
-
"seed": 42,
|
| 6 |
-
"epochs": 1000,
|
| 7 |
-
"learning_rate": 0.0001,
|
| 8 |
-
"betas": [
|
| 9 |
-
0.8,
|
| 10 |
-
0.99
|
| 11 |
-
],
|
| 12 |
-
"eps": 1e-09,
|
| 13 |
-
"batch_size": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/Azusa/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/Azusa/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"Azusa": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/Azusa/models/G_11300.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0708043b54ab21eb8ec1b600982ea7b105bcded370a9207281e043c64e195dc3
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/Carol/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 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": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/Carol/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/Carol/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"珈乐": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/Carol/models/G_12800.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:d95ce470d428646e70a3ef6b9f39bcdbabd0d60e5311c136acfa82c14140aa45
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/Diana/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 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": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/Diana/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/Diana/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"嘉然": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/Diana/models/G_7800.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:e13531f291fb5c277fbe2dc61674ce174cffd28cd228b40a0136a2c407fd2286
|
| 3 |
-
size 728370270
|
|
|
|
|
|
|
|
|
|
|
|
Data/Eileen/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 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": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/Eileen/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/Eileen/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"乃琳": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/Eileen/models/G_10700.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:905de797a2fb9382f22f160407f0dbf3dc4a24a98052dd3ab062bf0e13e58659
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/Eileen/models/G_10800.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:cb4607096aec15bbd73b7570cf25ad87ae56c13880398b6ced0aa57ee77866e7
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/Messi/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 50,
|
| 4 |
-
"eval_interval": 50,
|
| 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": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/Messi/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/Messi/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"Messi": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/Messi/models/G_1800.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:a812c363cd2e091a5126c33a2c8572fe06ce4b1840270d5c093e8ab7eba1b0ff
|
| 3 |
-
size 728370270
|
|
|
|
|
|
|
|
|
|
|
|
Data/Nana7mi/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 5 |
-
"seed": 42,
|
| 6 |
-
"epochs": 1000,
|
| 7 |
-
"learning_rate": 0.0001,
|
| 8 |
-
"betas": [
|
| 9 |
-
0.8,
|
| 10 |
-
0.99
|
| 11 |
-
],
|
| 12 |
-
"eps": 1e-09,
|
| 13 |
-
"batch_size": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/Nana7mi/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/Nana7mi/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"七海": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/Nana7mi/models/G_10700.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:616e10795efb68b46b3da14283b33dd11134c365a7866f596fde7f8654cd8a40
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/Taffy/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 5 |
-
"seed": 42,
|
| 6 |
-
"epochs": 1000,
|
| 7 |
-
"learning_rate": 0.0001,
|
| 8 |
-
"betas": [
|
| 9 |
-
0.8,
|
| 10 |
-
0.99
|
| 11 |
-
],
|
| 12 |
-
"eps": 1e-09,
|
| 13 |
-
"batch_size": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/Taffy/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/Taffy/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"永雏塔菲": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/Taffy/models/G_11100.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:39c4f4a231902fea06da77789a7949cbcdaaa26abbbb455c04760d2abf1eebea
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/XingTong/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 5 |
-
"seed": 42,
|
| 6 |
-
"epochs": 1000,
|
| 7 |
-
"learning_rate": 0.0001,
|
| 8 |
-
"betas": [
|
| 9 |
-
0.8,
|
| 10 |
-
0.99
|
| 11 |
-
],
|
| 12 |
-
"eps": 1e-09,
|
| 13 |
-
"batch_size": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/XingTong/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/XingTong/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"星瞳": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/XingTong/models/G_11600.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:1bf5588133ea19d099ffd762cd92601210ded38abc3ace35226dab333d81a436
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/badXT/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 50,
|
| 4 |
-
"eval_interval": 50,
|
| 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": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/badXT/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/badXT/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"坏女人星瞳": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/badXT/models/G_3250.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:70bccbfa08deb7ce10e8ab3d4d730b5a1dee4a5be6e326f03244aec4ea112658
|
| 3 |
-
size 728370270
|
|
|
|
|
|
|
|
|
|
|
|
Data/dingzhen/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 50,
|
| 4 |
-
"eval_interval": 50,
|
| 5 |
-
"seed": 42,
|
| 6 |
-
"epochs": 1000,
|
| 7 |
-
"learning_rate": 0.0001,
|
| 8 |
-
"betas": [
|
| 9 |
-
0.8,
|
| 10 |
-
0.99
|
| 11 |
-
],
|
| 12 |
-
"eps": 1e-09,
|
| 13 |
-
"batch_size": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/dingzhen/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/dingzhen/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"丁真": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/dingzhen/models/G_2650.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:540afcc29da68230b57c885fb00dd0dcd897b97f0796fab486ced3ede7fbfef7
|
| 3 |
-
size 728370270
|
|
|
|
|
|
|
|
|
|
|
|
Data/nine/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 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": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/nine/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/nine/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"尼奈": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/nine/models/G_10700.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:fe12e7d98b99bc9e26b45b9f191b1534a444ecddc630a313e6ea6439c5d486d3
|
| 3 |
-
size 728379830
|
|
|
|
|
|
|
|
|
|
|
|
Data/otto/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 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": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/otto/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/otto/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"otto": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 256,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/otto/models/G_5400.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:561e3ac21e3346e05f2b130a2833a28cd011d486ea14ff4a164e5f19094e5843
|
| 3 |
-
size 703791134
|
|
|
|
|
|
|
|
|
|
|
|
Data/xuan/config.json
DELETED
|
@@ -1,108 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"train": {
|
| 3 |
-
"log_interval": 100,
|
| 4 |
-
"eval_interval": 100,
|
| 5 |
-
"seed": 42,
|
| 6 |
-
"epochs": 1000,
|
| 7 |
-
"learning_rate": 0.0001,
|
| 8 |
-
"betas": [
|
| 9 |
-
0.8,
|
| 10 |
-
0.99
|
| 11 |
-
],
|
| 12 |
-
"eps": 1e-09,
|
| 13 |
-
"batch_size": 12,
|
| 14 |
-
"bf16_run": false,
|
| 15 |
-
"lr_decay": 0.99995,
|
| 16 |
-
"segment_size": 16384,
|
| 17 |
-
"init_lr_ratio": 1,
|
| 18 |
-
"warmup_epochs": 0,
|
| 19 |
-
"c_mel": 45,
|
| 20 |
-
"c_kl": 1.0,
|
| 21 |
-
"c_commit": 100,
|
| 22 |
-
"skip_optimizer": true,
|
| 23 |
-
"freeze_ZH_bert": false,
|
| 24 |
-
"freeze_JP_bert": false,
|
| 25 |
-
"freeze_EN_bert": false,
|
| 26 |
-
"freeze_emo": false
|
| 27 |
-
},
|
| 28 |
-
"data": {
|
| 29 |
-
"training_files": "Data/xuan/filelists/train.list",
|
| 30 |
-
"validation_files": "Data/xuan/filelists/val.list",
|
| 31 |
-
"max_wav_value": 32768.0,
|
| 32 |
-
"sampling_rate": 44100,
|
| 33 |
-
"filter_length": 2048,
|
| 34 |
-
"hop_length": 512,
|
| 35 |
-
"win_length": 2048,
|
| 36 |
-
"n_mel_channels": 128,
|
| 37 |
-
"mel_fmin": 0.0,
|
| 38 |
-
"mel_fmax": null,
|
| 39 |
-
"add_blank": true,
|
| 40 |
-
"n_speakers": 1,
|
| 41 |
-
"cleaned_text": true,
|
| 42 |
-
"spk2id": {
|
| 43 |
-
"炫神": 0
|
| 44 |
-
}
|
| 45 |
-
},
|
| 46 |
-
"model": {
|
| 47 |
-
"use_spk_conditioned_encoder": true,
|
| 48 |
-
"use_noise_scaled_mas": true,
|
| 49 |
-
"use_mel_posterior_encoder": false,
|
| 50 |
-
"use_duration_discriminator": true,
|
| 51 |
-
"inter_channels": 192,
|
| 52 |
-
"hidden_channels": 192,
|
| 53 |
-
"filter_channels": 768,
|
| 54 |
-
"n_heads": 2,
|
| 55 |
-
"n_layers": 6,
|
| 56 |
-
"kernel_size": 3,
|
| 57 |
-
"p_dropout": 0.1,
|
| 58 |
-
"resblock": "1",
|
| 59 |
-
"resblock_kernel_sizes": [
|
| 60 |
-
3,
|
| 61 |
-
7,
|
| 62 |
-
11
|
| 63 |
-
],
|
| 64 |
-
"resblock_dilation_sizes": [
|
| 65 |
-
[
|
| 66 |
-
1,
|
| 67 |
-
3,
|
| 68 |
-
5
|
| 69 |
-
],
|
| 70 |
-
[
|
| 71 |
-
1,
|
| 72 |
-
3,
|
| 73 |
-
5
|
| 74 |
-
],
|
| 75 |
-
[
|
| 76 |
-
1,
|
| 77 |
-
3,
|
| 78 |
-
5
|
| 79 |
-
]
|
| 80 |
-
],
|
| 81 |
-
"upsample_rates": [
|
| 82 |
-
8,
|
| 83 |
-
8,
|
| 84 |
-
2,
|
| 85 |
-
2,
|
| 86 |
-
2
|
| 87 |
-
],
|
| 88 |
-
"upsample_initial_channel": 512,
|
| 89 |
-
"upsample_kernel_sizes": [
|
| 90 |
-
16,
|
| 91 |
-
16,
|
| 92 |
-
8,
|
| 93 |
-
2,
|
| 94 |
-
2
|
| 95 |
-
],
|
| 96 |
-
"n_layers_q": 3,
|
| 97 |
-
"use_spectral_norm": false,
|
| 98 |
-
"gin_channels": 512,
|
| 99 |
-
"slm": {
|
| 100 |
-
"model": "./slm/wavlm-base-plus",
|
| 101 |
-
"sr": 16000,
|
| 102 |
-
"hidden": 768,
|
| 103 |
-
"nlayers": 13,
|
| 104 |
-
"initial_channel": 64
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"version": "2.3"
|
| 108 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Data/xuan/models/G_7200.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:7fd876c3eb726c3e15f1df3260a0821f84d955944044cca119e10d021dd58300
|
| 3 |
-
size 728370270
|
|
|
|
|
|
|
|
|
|
|
|