Upload config.json
Browse files- speechsr48k/config.json +48 -0
speechsr48k/config.json
ADDED
|
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
{
|
| 3 |
+
"train": {
|
| 4 |
+
"log_interval": 200,
|
| 5 |
+
"eval_interval": 10000,
|
| 6 |
+
"save_interval": 10000,
|
| 7 |
+
"seed": 1234,
|
| 8 |
+
"epochs": 20000,
|
| 9 |
+
"learning_rate": 1e-4,
|
| 10 |
+
"betas": [0.8, 0.99],
|
| 11 |
+
"eps": 1e-9,
|
| 12 |
+
"batch_size": 32,
|
| 13 |
+
"fp16_run": false,
|
| 14 |
+
"lr_decay": 0.995,
|
| 15 |
+
"segment_size": 9600,
|
| 16 |
+
"init_lr_ratio": 1,
|
| 17 |
+
"warmup_epochs": 0,
|
| 18 |
+
"c_mel": 45
|
| 19 |
+
},
|
| 20 |
+
"data": {
|
| 21 |
+
"train_filelist_path": "filelists/train_48k_vctk_trim_bigvgan_sr.txt",
|
| 22 |
+
"test_filelist_path": "filelists/test_48k_vctk_trim_bigvgan_sr.txt",
|
| 23 |
+
"text_cleaners":["english_cleaners2"],
|
| 24 |
+
"max_wav_value": 32768.0,
|
| 25 |
+
"sampling_rate": 48000,
|
| 26 |
+
"filter_length": 1280,
|
| 27 |
+
"hop_length": 320,
|
| 28 |
+
"win_length": 1280,
|
| 29 |
+
"n_mel_channels": 128,
|
| 30 |
+
"mel_fmin": 0,
|
| 31 |
+
"mel_fmax": 24000,
|
| 32 |
+
"add_blank": true,
|
| 33 |
+
"n_speakers": 0,
|
| 34 |
+
"cleaned_text": true,
|
| 35 |
+
"aug_rate": 1.0,
|
| 36 |
+
"top_db": 20
|
| 37 |
+
},
|
| 38 |
+
"model": {
|
| 39 |
+
"resblock": "0",
|
| 40 |
+
"resblock_kernel_sizes": [3,7,11],
|
| 41 |
+
"resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
|
| 42 |
+
"upsample_rates": [3],
|
| 43 |
+
"upsample_initial_channel": 32,
|
| 44 |
+
"upsample_kernel_sizes": [3],
|
| 45 |
+
"use_spectral_norm": false
|
| 46 |
+
}
|
| 47 |
+
}
|
| 48 |
+
|