Upload 4 files
Browse files- ckpt/config_bigvgan.json +54 -0
- ckpt/model_diffhier.pth +3 -0
- vocoder/voc_bigvgan.pth +3 -0
- vocoder/voc_hifigan.pth +3 -0
ckpt/config_bigvgan.json
ADDED
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"train": {
|
3 |
+
"log_interval": 1000,
|
4 |
+
"eval_interval": 10000,
|
5 |
+
"save_interval": 10000,
|
6 |
+
"seed": 1234,
|
7 |
+
"epochs": 1000,
|
8 |
+
"optimizer": "adamw",
|
9 |
+
"lr_decay_on": true,
|
10 |
+
"learning_rate": 5e-5,
|
11 |
+
"betas": [0.8, 0.99],
|
12 |
+
"eps": 1e-9,
|
13 |
+
"batch_size": 32,
|
14 |
+
"fp16_run": false,
|
15 |
+
"lr_decay": 0.999875,
|
16 |
+
"segment_size": 35840,
|
17 |
+
"init_lr_ratio": 1,
|
18 |
+
"warmup_epochs": 0,
|
19 |
+
"c_mel": 1,
|
20 |
+
"aug": true,
|
21 |
+
"lambda_commit": 0.02
|
22 |
+
},
|
23 |
+
"data": {
|
24 |
+
"sampling_rate": 16000,
|
25 |
+
"filter_length": 1280,
|
26 |
+
"hop_length": 320,
|
27 |
+
"win_length": 1280,
|
28 |
+
"n_mel_channels": 80,
|
29 |
+
"mel_fmin": 0,
|
30 |
+
"mel_fmax": 8000
|
31 |
+
},
|
32 |
+
"model": {
|
33 |
+
"inter_channels": 192,
|
34 |
+
"hidden_channels": 192,
|
35 |
+
"filter_channels": 768,
|
36 |
+
"n_heads": 2,
|
37 |
+
"n_layers": 8,
|
38 |
+
"kernel_size": 3,
|
39 |
+
"p_dropout": 0.1,
|
40 |
+
"resblock": "1",
|
41 |
+
"resblock_kernel_sizes": [3,7,11],
|
42 |
+
"resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
|
43 |
+
"upsample_rates": [5,4,2,2,2,2],
|
44 |
+
"upsample_initial_channel": 1024,
|
45 |
+
"upsample_kernel_sizes": [11,8,4,4,4,4],
|
46 |
+
"hidden_size": 128
|
47 |
+
},
|
48 |
+
"diffusion" : {
|
49 |
+
"dec_dim" : 64,
|
50 |
+
"spk_dim" : 128,
|
51 |
+
"beta_min" : 0.05,
|
52 |
+
"beta_max" : 20.0
|
53 |
+
}
|
54 |
+
}
|
ckpt/model_diffhier.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:aab3e6972ba6bbc4b3f5bf032ab6375daf00945221abbcf590aaafa362dc7875
|
3 |
+
size 73382935
|
vocoder/voc_bigvgan.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7169372d42c08aec46cdc639cdaa29cd2faa722138c477f843d4d7f1ebc22cb4
|
3 |
+
size 620828487
|
vocoder/voc_hifigan.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1c8e7b8a55739edf6845ef37f4af4efa4af4f6bbf930842c28115a283ed3ee06
|
3 |
+
size 157373725
|