litagin commited on
Commit
db4789e
1 Parent(s): c18724c

Upload 3 files

Browse files
minus_flat_null/config.json ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "minus_flat_null",
3
+ "train": {
4
+ "log_interval": 200,
5
+ "eval_interval": 1000,
6
+ "seed": 42,
7
+ "epochs": 100,
8
+ "learning_rate": 0.0001,
9
+ "betas": [0.8, 0.99],
10
+ "eps": 1e-9,
11
+ "batch_size": 4,
12
+ "bf16_run": false,
13
+ "fp16_run": false,
14
+ "lr_decay": 0.99996,
15
+ "segment_size": 16384,
16
+ "init_lr_ratio": 1,
17
+ "warmup_epochs": 0,
18
+ "c_mel": 45,
19
+ "c_kl": 1.0,
20
+ "c_commit": 100,
21
+ "skip_optimizer": false,
22
+ "freeze_ZH_bert": false,
23
+ "freeze_JP_bert": false,
24
+ "freeze_EN_bert": false,
25
+ "freeze_emo": false,
26
+ "freeze_style": false,
27
+ "freeze_decoder": false
28
+ },
29
+ "data": {
30
+ "use_jp_extra": true,
31
+ "training_files": "Data\\Dummy\\train.list",
32
+ "validation_files": "Data\\Dummy\\val.list",
33
+ "max_wav_value": 32768.0,
34
+ "sampling_rate": 44100,
35
+ "filter_length": 2048,
36
+ "hop_length": 512,
37
+ "win_length": 2048,
38
+ "n_mel_channels": 128,
39
+ "mel_fmin": 0.0,
40
+ "mel_fmax": null,
41
+ "add_blank": true,
42
+ "n_speakers": 1,
43
+ "cleaned_text": true,
44
+ "spk2id": {
45
+ "Dummy": 0
46
+ },
47
+ "num_styles": 1,
48
+ "style2id": {
49
+ "Neutral": 0
50
+ }
51
+ },
52
+ "model": {
53
+ "use_spk_conditioned_encoder": true,
54
+ "use_noise_scaled_mas": true,
55
+ "use_mel_posterior_encoder": false,
56
+ "use_duration_discriminator": false,
57
+ "use_wavlm_discriminator": true,
58
+ "inter_channels": 192,
59
+ "hidden_channels": 192,
60
+ "filter_channels": 768,
61
+ "n_heads": 2,
62
+ "n_layers": 6,
63
+ "kernel_size": 3,
64
+ "p_dropout": 0.1,
65
+ "resblock": "1",
66
+ "resblock_kernel_sizes": [3, 7, 11],
67
+ "resblock_dilation_sizes": [
68
+ [1, 3, 5],
69
+ [1, 3, 5],
70
+ [1, 3, 5]
71
+ ],
72
+ "upsample_rates": [8, 8, 2, 2, 2],
73
+ "upsample_initial_channel": 512,
74
+ "upsample_kernel_sizes": [16, 16, 8, 2, 2],
75
+ "n_layers_q": 3,
76
+ "use_spectral_norm": false,
77
+ "gin_channels": 512,
78
+ "slm": {
79
+ "model": "./slm/wavlm-base-plus",
80
+ "sr": 16000,
81
+ "hidden": 768,
82
+ "nlayers": 13,
83
+ "initial_channel": 64
84
+ }
85
+ },
86
+ "version": "2.6.0-JP-Extra"
87
+ }
minus_flat_null/minus_flat_null.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:976016eb37af706946f40302d238658d125072e1b7cafdd65954118e7c0487c5
3
+ size 251150980
minus_flat_null/style_vectors.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3855acfe4b6e19f936afa491c0daac46a72ba91d05924492483a8763b74d3890
3
+ size 1152