Create s2D2333k.json
Browse files
gsv-v2final-pretrained/s2D2333k.json
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
'train': {
|
| 3 |
+
'log_interval': 100,
|
| 4 |
+
'eval_interval': 200,
|
| 5 |
+
'seed': 1234,
|
| 6 |
+
'epochs': 15,
|
| 7 |
+
'learning_rate': 0.0001,
|
| 8 |
+
'betas': [
|
| 9 |
+
0.8,
|
| 10 |
+
0.99
|
| 11 |
+
],
|
| 12 |
+
'eps': 1e-09,
|
| 13 |
+
'batch_size': 8,
|
| 14 |
+
'fp16_run': False,
|
| 15 |
+
'lr_decay': 0.999875,
|
| 16 |
+
'segment_size': 20480,
|
| 17 |
+
'init_lr_ratio': 1,
|
| 18 |
+
'warmup_epochs': 0,
|
| 19 |
+
'c_mel': 45,
|
| 20 |
+
'c_kl': 1.0
|
| 21 |
+
},
|
| 22 |
+
'data': {
|
| 23 |
+
'max_wav_value': 32768.0,
|
| 24 |
+
'sampling_rate': 32000,
|
| 25 |
+
'filter_length': 2048,
|
| 26 |
+
'hop_length': 640,
|
| 27 |
+
'win_length': 2048,
|
| 28 |
+
'n_mel_channels': 128,
|
| 29 |
+
'mel_fmin': 0.0,
|
| 30 |
+
'mel_fmax': None,
|
| 31 |
+
'add_blank': True,
|
| 32 |
+
'n_speakers': 300,
|
| 33 |
+
'cleaned_text': True
|
| 34 |
+
},
|
| 35 |
+
'model': {
|
| 36 |
+
'inter_channels': 192,
|
| 37 |
+
'hidden_channels': 192,
|
| 38 |
+
'filter_channels': 768,
|
| 39 |
+
'n_heads': 2,
|
| 40 |
+
'n_layers': 6,
|
| 41 |
+
'kernel_size': 3,
|
| 42 |
+
'p_dropout': 0.1,
|
| 43 |
+
'resblock': '1',
|
| 44 |
+
'resblock_kernel_sizes': [
|
| 45 |
+
3,
|
| 46 |
+
7,
|
| 47 |
+
11
|
| 48 |
+
],
|
| 49 |
+
'resblock_dilation_sizes': [
|
| 50 |
+
[
|
| 51 |
+
1,
|
| 52 |
+
3,
|
| 53 |
+
5
|
| 54 |
+
],
|
| 55 |
+
[
|
| 56 |
+
1,
|
| 57 |
+
3,
|
| 58 |
+
5
|
| 59 |
+
],
|
| 60 |
+
[
|
| 61 |
+
1,
|
| 62 |
+
3,
|
| 63 |
+
5
|
| 64 |
+
]
|
| 65 |
+
],
|
| 66 |
+
'upsample_rates': [
|
| 67 |
+
10,
|
| 68 |
+
8,
|
| 69 |
+
2,
|
| 70 |
+
2,
|
| 71 |
+
2
|
| 72 |
+
],
|
| 73 |
+
'upsample_initial_channel': 512,
|
| 74 |
+
'upsample_kernel_sizes': [
|
| 75 |
+
16,
|
| 76 |
+
16,
|
| 77 |
+
8,
|
| 78 |
+
2,
|
| 79 |
+
2
|
| 80 |
+
],
|
| 81 |
+
'n_layers_q': 3,
|
| 82 |
+
'use_spectral_norm': False,
|
| 83 |
+
'gin_channels': 512,
|
| 84 |
+
'semantic_frame_rate': '25hz',
|
| 85 |
+
'freeze_quantizer': True
|
| 86 |
+
},
|
| 87 |
+
's2_ckpt_dir': 'logs/s2/big2k1ft_ko_5commit2',
|
| 88 |
+
'content_module': 'cnhubert'
|
| 89 |
+
}
|