Upload config.json with huggingface_hub
Browse files- config.json +105 -0
config.json
ADDED
|
@@ -0,0 +1,105 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"train": {
|
| 3 |
+
"log_interval": 200,
|
| 4 |
+
"eval_interval": 1000,
|
| 5 |
+
"seed": 52,
|
| 6 |
+
"epochs": 200,
|
| 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": true,
|
| 15 |
+
"lr_decay": 0.999,
|
| 16 |
+
"segment_size": 16384,
|
| 17 |
+
"init_lr_ratio": 1,
|
| 18 |
+
"warmup_epochs": 0,
|
| 19 |
+
"c_mel": 45,
|
| 20 |
+
"c_kl": 1.0,
|
| 21 |
+
"skip_optimizer": true,
|
| 22 |
+
"keep_ckpts": 5
|
| 23 |
+
},
|
| 24 |
+
"data": {
|
| 25 |
+
"training_files": "/home/hoangvanson/vietnamese_tts_standalone/english_tts_standalone/processed_data_english/train_ljspeech.list",
|
| 26 |
+
"validation_files": "/home/hoangvanson/vietnamese_tts_standalone/english_tts_standalone/processed_data_english/val_ljspeech.list",
|
| 27 |
+
"max_wav_value": 32768.0,
|
| 28 |
+
"sampling_rate": 44100,
|
| 29 |
+
"filter_length": 2048,
|
| 30 |
+
"hop_length": 512,
|
| 31 |
+
"win_length": 2048,
|
| 32 |
+
"n_mel_channels": 128,
|
| 33 |
+
"mel_fmin": 0.0,
|
| 34 |
+
"mel_fmax": null,
|
| 35 |
+
"add_blank": true,
|
| 36 |
+
"n_speakers": 1,
|
| 37 |
+
"use_speaker_encoder": false,
|
| 38 |
+
"speaker_embed_dim": 128,
|
| 39 |
+
"cleaned_text": true,
|
| 40 |
+
"spk2id": {
|
| 41 |
+
"MALE": 0
|
| 42 |
+
},
|
| 43 |
+
"disable_bert": true
|
| 44 |
+
},
|
| 45 |
+
"model": {
|
| 46 |
+
"use_spk_conditioned_encoder": true,
|
| 47 |
+
"use_noise_scaled_mas": true,
|
| 48 |
+
"use_mel_posterior_encoder": false,
|
| 49 |
+
"use_duration_discriminator": true,
|
| 50 |
+
"inter_channels": 32,
|
| 51 |
+
"hidden_channels": 32,
|
| 52 |
+
"filter_channels": 128,
|
| 53 |
+
"n_heads": 2,
|
| 54 |
+
"n_layers": 3,
|
| 55 |
+
"n_layers_trans_flow": 3,
|
| 56 |
+
"kernel_size": 3,
|
| 57 |
+
"p_dropout": 0.1,
|
| 58 |
+
"resblock": "1",
|
| 59 |
+
"resblock_kernel_sizes": [
|
| 60 |
+
3,
|
| 61 |
+
7,
|
| 62 |
+
11
|
| 63 |
+
],
|
| 64 |
+
"resblock_dilation_sizes": [
|
| 65 |
+
[
|
| 66 |
+
1,
|
| 67 |
+
3,
|
| 68 |
+
5
|
| 69 |
+
],
|
| 70 |
+
[
|
| 71 |
+
1,
|
| 72 |
+
3,
|
| 73 |
+
5
|
| 74 |
+
],
|
| 75 |
+
[
|
| 76 |
+
1,
|
| 77 |
+
3,
|
| 78 |
+
5
|
| 79 |
+
]
|
| 80 |
+
],
|
| 81 |
+
"upsample_rates": [
|
| 82 |
+
8,
|
| 83 |
+
8,
|
| 84 |
+
2,
|
| 85 |
+
2,
|
| 86 |
+
2
|
| 87 |
+
],
|
| 88 |
+
"upsample_initial_channel": 64,
|
| 89 |
+
"upsample_kernel_sizes": [
|
| 90 |
+
16,
|
| 91 |
+
16,
|
| 92 |
+
8,
|
| 93 |
+
2,
|
| 94 |
+
2
|
| 95 |
+
],
|
| 96 |
+
"n_layers_q": 3,
|
| 97 |
+
"use_spectral_norm": false,
|
| 98 |
+
"gin_channels": 128,
|
| 99 |
+
"use_sdp": true,
|
| 100 |
+
"mas_noise_scale_initial": 0.01,
|
| 101 |
+
"noise_scale_delta": 2e-06
|
| 102 |
+
},
|
| 103 |
+
"num_languages": 1,
|
| 104 |
+
"num_tones": 6
|
| 105 |
+
}
|