frappuccino commited on
Commit
aca3797
1 Parent(s): fe1f9d3

Upload 3 files

Browse files
Files changed (3) hide show
  1. DUR_138000.pth +3 -0
  2. G_138000.pth +3 -0
  3. config.json +61 -0
DUR_138000.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f894c42b4db61a6eb971f79857f4683af7ac4870e1eac17fa5bb783373f286e
3
+ size 6670315
G_138000.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8367f47cca6acde645abad93ebf656439b352f4d9ddd39cf61a1d01a069aae03
3
+ size 565452507
config.json ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "log_interval": 200,
4
+ "eval_interval": 1000,
5
+ "seed": 1234,
6
+ "epochs": 20000,
7
+ "learning_rate": 1e-4,
8
+ "betas": [0.8, 0.99],
9
+ "eps": 1e-9,
10
+ "batch_size": 16,
11
+ "fp16_run": false,
12
+ "lr_decay": 0.999875,
13
+ "segment_size": 8192,
14
+ "init_lr_ratio": 1,
15
+ "warmup_epochs": 0,
16
+ "c_mel": 45,
17
+ "c_kl": 1.0
18
+ },
19
+ "data": {
20
+ "use_mel_posterior_encoder": true,
21
+ "training_files":"/media/kingspec/Datasets/Russian/Natasha/metadata_fullname_train.csv.cleaned",
22
+ "validation_files":"/media/kingspec/Datasets/Russian/Natasha/metadata_fullname_val.csv.cleaned",
23
+ "text_cleaners":["basic_cleaners"],
24
+ "max_wav_value": 32768.0,
25
+ "sampling_rate": 22050,
26
+ "filter_length": 1024,
27
+ "hop_length": 256,
28
+ "win_length": 1024,
29
+ "n_mel_channels": 80,
30
+ "mel_fmin": 0.0,
31
+ "mel_fmax": null,
32
+ "add_blank": false,
33
+ "n_speakers": 0,
34
+ "cleaned_text": true
35
+ },
36
+ "model": {
37
+ "use_mel_posterior_encoder": true,
38
+ "use_transformer_flows": true,
39
+ "transformer_flow_type": "pre_conv",
40
+ "use_spk_conditioned_encoder": false,
41
+ "use_noise_scaled_mas": true,
42
+ "use_duration_discriminator": true,
43
+ "inter_channels": 192,
44
+ "hidden_channels": 192,
45
+ "filter_channels": 768,
46
+ "n_heads": 2,
47
+ "n_layers": 10,
48
+ "kernel_size": 5,
49
+ "p_dropout": 0.1,
50
+ "resblock": "1",
51
+ "resblock_kernel_sizes": [3,7,11],
52
+ "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
53
+ "upsample_rates": [8,8,2,2],
54
+ "upsample_initial_channel": 512,
55
+ "upsample_kernel_sizes": [16,16,4,4],
56
+ "n_layers_q": 3,
57
+ "use_spectral_norm": false,
58
+ "use_sdp": false
59
+ }
60
+ }
61
+