lpw commited on
Commit
82b1682
1 Parent(s): fea0313

Upload 2 files

Browse files
Files changed (2) hide show
  1. config.json +55 -0
  2. model.pt +3 -0
config.json ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "input_wavs_dir": "/private/home/adampolyak/datasets/LJ/LJSpeech-1.1/wavs_16khz_padded",
3
+ "input_training_file": "/large_experiments/ust/annl/datasets/tts/TAT/filelist/Zh_Hokkien_pretrain_it3_alpha0.05_400k_it2_400k_it1_400k/layer12.km2500_frame_TAT-TTS-all_train.txt",
4
+ "input_validation_file": "/large_experiments/ust/annl/datasets/tts/TAT/filelist/Zh_Hokkien_pretrain_it3_alpha0.05_400k_it2_400k_it1_400k/layer12.km2500_frame_TAT-TTS-all_dev.txt",
5
+
6
+ "resblock": "1",
7
+ "num_gpus": 0,
8
+ "batch_size": 16,
9
+ "learning_rate": 0.0002,
10
+ "adam_b1": 0.8,
11
+ "adam_b2": 0.99,
12
+ "lr_decay": 0.999,
13
+ "seed": 1234,
14
+
15
+ "upsample_rates": [5,4,4,2,2],
16
+ "upsample_kernel_sizes": [11,8,8,4,4],
17
+ "upsample_initial_channel": 512,
18
+ "resblock_kernel_sizes": [3,7,11],
19
+ "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
20
+ "num_embeddings": 2500,
21
+ "embedding_dim": 128,
22
+ "model_in_dim": 256,
23
+ "multispkr": "_",
24
+
25
+ "segment_size": 8960,
26
+ "code_hop_size": 320,
27
+ "f0": false,
28
+ "num_mels": 80,
29
+ "num_freq": 1025,
30
+ "n_fft": 1024,
31
+ "hop_size": 256,
32
+ "win_size": 1024,
33
+
34
+
35
+ "dur_prediction_weight": 1.0,
36
+ "dur_predictor_params": {
37
+ "encoder_embed_dim": 128,
38
+ "var_pred_hidden_dim": 128,
39
+ "var_pred_kernel_size": 3,
40
+ "var_pred_dropout": 0.5
41
+ },
42
+
43
+ "sampling_rate": 16000,
44
+
45
+ "fmin": 0,
46
+ "fmax": 8000,
47
+ "fmax_for_loss": null,
48
+
49
+ "num_workers": 4,
50
+
51
+ "dist_config": {
52
+ "dist_backend": "nccl",
53
+ "dist_url": "env://"
54
+ }
55
+ }
model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3210a22eeced7d200c2f950a047151a4f554e44d0eb581836c670cb69d8802cd
3
+ size 56756950