Commit
•
fff9519
1
Parent(s):
3dc3176
upload: model pth and configs
Browse files- config.json +761 -0
- language_ids.json +4 -0
- model.pth +3 -0
- speakers.pth +3 -0
config.json
ADDED
|
@@ -0,0 +1,761 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"output_path": "/mnt/user_wayne/TTS/recipes/hat-tts/vits",
|
| 3 |
+
"logger_uri": null,
|
| 4 |
+
"run_name": "vits_hat_tts",
|
| 5 |
+
"project_name": "vits_hat_tts",
|
| 6 |
+
"run_description": "\ud83d\udc38Coqui trainer run.",
|
| 7 |
+
"print_step": 50,
|
| 8 |
+
"plot_step": 100,
|
| 9 |
+
"model_param_stats": false,
|
| 10 |
+
"wandb_entity": null,
|
| 11 |
+
"dashboard_logger": "wandb",
|
| 12 |
+
"save_on_interrupt": true,
|
| 13 |
+
"log_model_step": 10000,
|
| 14 |
+
"save_step": 10000,
|
| 15 |
+
"save_n_checkpoints": 5,
|
| 16 |
+
"save_checkpoints": true,
|
| 17 |
+
"save_all_best": false,
|
| 18 |
+
"save_best_after": 0,
|
| 19 |
+
"target_loss": null,
|
| 20 |
+
"print_eval": true,
|
| 21 |
+
"test_delay_epochs": -1,
|
| 22 |
+
"run_eval": true,
|
| 23 |
+
"run_eval_steps": null,
|
| 24 |
+
"distributed_backend": "nccl",
|
| 25 |
+
"distributed_url": "tcp://localhost:54321",
|
| 26 |
+
"mixed_precision": true,
|
| 27 |
+
"precision": "fp16",
|
| 28 |
+
"epochs": 1000,
|
| 29 |
+
"batch_size": 32,
|
| 30 |
+
"eval_batch_size": 32,
|
| 31 |
+
"grad_clip": [
|
| 32 |
+
1000,
|
| 33 |
+
1000
|
| 34 |
+
],
|
| 35 |
+
"scheduler_after_epoch": true,
|
| 36 |
+
"lr": 0.001,
|
| 37 |
+
"optimizer": "AdamW",
|
| 38 |
+
"optimizer_params": {
|
| 39 |
+
"betas": [
|
| 40 |
+
0.8,
|
| 41 |
+
0.99
|
| 42 |
+
],
|
| 43 |
+
"eps": 1e-09,
|
| 44 |
+
"weight_decay": 0.01
|
| 45 |
+
},
|
| 46 |
+
"lr_scheduler": null,
|
| 47 |
+
"lr_scheduler_params": {},
|
| 48 |
+
"use_grad_scaler": false,
|
| 49 |
+
"allow_tf32": false,
|
| 50 |
+
"cudnn_enable": true,
|
| 51 |
+
"cudnn_deterministic": false,
|
| 52 |
+
"cudnn_benchmark": false,
|
| 53 |
+
"training_seed": 54321,
|
| 54 |
+
"model": "vits",
|
| 55 |
+
"num_loader_workers": 16,
|
| 56 |
+
"num_eval_loader_workers": 16,
|
| 57 |
+
"use_noise_augment": false,
|
| 58 |
+
"audio": {
|
| 59 |
+
"fft_size": 1024,
|
| 60 |
+
"sample_rate": 16000,
|
| 61 |
+
"win_length": 1024,
|
| 62 |
+
"hop_length": 256,
|
| 63 |
+
"num_mels": 80,
|
| 64 |
+
"mel_fmin": 0,
|
| 65 |
+
"mel_fmax": null
|
| 66 |
+
},
|
| 67 |
+
"use_phonemes": false,
|
| 68 |
+
"phonemizer": null,
|
| 69 |
+
"phoneme_language": null,
|
| 70 |
+
"compute_input_seq_cache": true,
|
| 71 |
+
"text_cleaner": null,
|
| 72 |
+
"enable_eos_bos_chars": false,
|
| 73 |
+
"test_sentences_file": "",
|
| 74 |
+
"phoneme_cache_path": null,
|
| 75 |
+
"characters": {
|
| 76 |
+
"characters_class": "TTS.tts.models.vits.VitsCharacters",
|
| 77 |
+
"vocab_dict": null,
|
| 78 |
+
"pad": "<PAD>",
|
| 79 |
+
"eos": "<EOS>",
|
| 80 |
+
"bos": "<BOS>",
|
| 81 |
+
"blank": "<BLNK>",
|
| 82 |
+
"characters": [
|
| 83 |
+
"a",
|
| 84 |
+
"e",
|
| 85 |
+
"f",
|
| 86 |
+
"h",
|
| 87 |
+
"i",
|
| 88 |
+
"k",
|
| 89 |
+
"l",
|
| 90 |
+
"m",
|
| 91 |
+
"n",
|
| 92 |
+
"o",
|
| 93 |
+
"p",
|
| 94 |
+
"s",
|
| 95 |
+
"t",
|
| 96 |
+
"u",
|
| 97 |
+
"v",
|
| 98 |
+
"\u014b",
|
| 99 |
+
"\u0255",
|
| 100 |
+
"\u0259",
|
| 101 |
+
"\u0268",
|
| 102 |
+
"\u0282",
|
| 103 |
+
"\u0288",
|
| 104 |
+
"\u0290",
|
| 105 |
+
"\u02b0",
|
| 106 |
+
"\u0329",
|
| 107 |
+
"\u0361",
|
| 108 |
+
"\uff0c",
|
| 109 |
+
"11",
|
| 110 |
+
"2",
|
| 111 |
+
"24",
|
| 112 |
+
"31",
|
| 113 |
+
"33",
|
| 114 |
+
"5",
|
| 115 |
+
"53",
|
| 116 |
+
"55"
|
| 117 |
+
],
|
| 118 |
+
"punctuations": "\uff0c ",
|
| 119 |
+
"phonemes": null,
|
| 120 |
+
"is_unique": true,
|
| 121 |
+
"is_sorted": true
|
| 122 |
+
},
|
| 123 |
+
"add_blank": true,
|
| 124 |
+
"batch_group_size": 32,
|
| 125 |
+
"loss_masking": null,
|
| 126 |
+
"min_audio_len": 1,
|
| 127 |
+
"max_audio_len": 1440000,
|
| 128 |
+
"min_text_len": 1,
|
| 129 |
+
"max_text_len": Infinity,
|
| 130 |
+
"compute_f0": false,
|
| 131 |
+
"compute_energy": false,
|
| 132 |
+
"compute_linear_spec": true,
|
| 133 |
+
"precompute_num_workers": 0,
|
| 134 |
+
"start_by_longest": false,
|
| 135 |
+
"shuffle": false,
|
| 136 |
+
"drop_last": false,
|
| 137 |
+
"datasets": [
|
| 138 |
+
{
|
| 139 |
+
"formatter": "hat_tts",
|
| 140 |
+
"dataset_name": "",
|
| 141 |
+
"path": "/mnt/user_wayne/hat_tts",
|
| 142 |
+
"meta_file_train": "sixian.json",
|
| 143 |
+
"ignored_speakers": null,
|
| 144 |
+
"language": "sixian",
|
| 145 |
+
"phonemizer": "",
|
| 146 |
+
"meta_file_val": "",
|
| 147 |
+
"meta_file_attn_mask": ""
|
| 148 |
+
},
|
| 149 |
+
{
|
| 150 |
+
"formatter": "hat_tts",
|
| 151 |
+
"dataset_name": "",
|
| 152 |
+
"path": "/mnt/user_wayne/hat_tts",
|
| 153 |
+
"meta_file_train": "hailu.json",
|
| 154 |
+
"ignored_speakers": null,
|
| 155 |
+
"language": "hailu",
|
| 156 |
+
"phonemizer": "",
|
| 157 |
+
"meta_file_val": "",
|
| 158 |
+
"meta_file_attn_mask": ""
|
| 159 |
+
}
|
| 160 |
+
],
|
| 161 |
+
"test_sentences": [
|
| 162 |
+
[
|
| 163 |
+
[
|
| 164 |
+
"a",
|
| 165 |
+
"n",
|
| 166 |
+
"31",
|
| 167 |
+
"t",
|
| 168 |
+
"\u0361",
|
| 169 |
+
"s",
|
| 170 |
+
"o",
|
| 171 |
+
"31",
|
| 172 |
+
" ",
|
| 173 |
+
"\uff0c",
|
| 174 |
+
" ",
|
| 175 |
+
"s",
|
| 176 |
+
"\u0268",
|
| 177 |
+
"t",
|
| 178 |
+
"5",
|
| 179 |
+
"p",
|
| 180 |
+
"a",
|
| 181 |
+
"u",
|
| 182 |
+
"31",
|
| 183 |
+
"m",
|
| 184 |
+
"a",
|
| 185 |
+
"\u014b",
|
| 186 |
+
"11",
|
| 187 |
+
" ",
|
| 188 |
+
"\uff0c",
|
| 189 |
+
" ",
|
| 190 |
+
"s",
|
| 191 |
+
"\u0268",
|
| 192 |
+
"t",
|
| 193 |
+
"5",
|
| 194 |
+
"p",
|
| 195 |
+
"a",
|
| 196 |
+
"u",
|
| 197 |
+
"31",
|
| 198 |
+
" ",
|
| 199 |
+
"l",
|
| 200 |
+
"e",
|
| 201 |
+
"24",
|
| 202 |
+
" ",
|
| 203 |
+
"\uff0c",
|
| 204 |
+
" ",
|
| 205 |
+
"n",
|
| 206 |
+
"\u0329",
|
| 207 |
+
"11",
|
| 208 |
+
" ",
|
| 209 |
+
"\u014b",
|
| 210 |
+
"a",
|
| 211 |
+
"i",
|
| 212 |
+
"11",
|
| 213 |
+
" ",
|
| 214 |
+
"k",
|
| 215 |
+
"i",
|
| 216 |
+
"11",
|
| 217 |
+
" ",
|
| 218 |
+
"\uff0c",
|
| 219 |
+
" ",
|
| 220 |
+
"\u014b",
|
| 221 |
+
"\u0329",
|
| 222 |
+
"11",
|
| 223 |
+
"k",
|
| 224 |
+
"o",
|
| 225 |
+
"\u014b",
|
| 226 |
+
"31",
|
| 227 |
+
"m",
|
| 228 |
+
"a",
|
| 229 |
+
"31",
|
| 230 |
+
"k",
|
| 231 |
+
"e",
|
| 232 |
+
"55",
|
| 233 |
+
" ",
|
| 234 |
+
"\uff0c",
|
| 235 |
+
" ",
|
| 236 |
+
"n",
|
| 237 |
+
"\u0329",
|
| 238 |
+
"11",
|
| 239 |
+
" ",
|
| 240 |
+
"h",
|
| 241 |
+
"e",
|
| 242 |
+
"55",
|
| 243 |
+
"m",
|
| 244 |
+
"o",
|
| 245 |
+
"11",
|
| 246 |
+
" ",
|
| 247 |
+
"s",
|
| 248 |
+
"a",
|
| 249 |
+
"11",
|
| 250 |
+
" ",
|
| 251 |
+
"\uff0c",
|
| 252 |
+
" ",
|
| 253 |
+
"\u014b",
|
| 254 |
+
"a",
|
| 255 |
+
"i",
|
| 256 |
+
"11",
|
| 257 |
+
" ",
|
| 258 |
+
"h",
|
| 259 |
+
"e",
|
| 260 |
+
"55",
|
| 261 |
+
" ",
|
| 262 |
+
"a",
|
| 263 |
+
"24",
|
| 264 |
+
" ",
|
| 265 |
+
"l",
|
| 266 |
+
"i",
|
| 267 |
+
"n",
|
| 268 |
+
"11",
|
| 269 |
+
" ",
|
| 270 |
+
"\uff0c",
|
| 271 |
+
" ",
|
| 272 |
+
"a",
|
| 273 |
+
"n",
|
| 274 |
+
"31",
|
| 275 |
+
"k",
|
| 276 |
+
"i",
|
| 277 |
+
"u",
|
| 278 |
+
"31",
|
| 279 |
+
"m",
|
| 280 |
+
"o",
|
| 281 |
+
"11",
|
| 282 |
+
"k",
|
| 283 |
+
"\u02b0",
|
| 284 |
+
"o",
|
| 285 |
+
"n",
|
| 286 |
+
"55",
|
| 287 |
+
"t",
|
| 288 |
+
"o",
|
| 289 |
+
"31",
|
| 290 |
+
" ",
|
| 291 |
+
"\uff0c",
|
| 292 |
+
" ",
|
| 293 |
+
"\u014b",
|
| 294 |
+
"i",
|
| 295 |
+
"11",
|
| 296 |
+
"h",
|
| 297 |
+
"o",
|
| 298 |
+
"31",
|
| 299 |
+
" ",
|
| 300 |
+
"m",
|
| 301 |
+
"o",
|
| 302 |
+
"11"
|
| 303 |
+
],
|
| 304 |
+
"XF",
|
| 305 |
+
null,
|
| 306 |
+
"sixian"
|
| 307 |
+
],
|
| 308 |
+
[
|
| 309 |
+
[
|
| 310 |
+
"a",
|
| 311 |
+
"n",
|
| 312 |
+
"31",
|
| 313 |
+
"t",
|
| 314 |
+
"\u0361",
|
| 315 |
+
"s",
|
| 316 |
+
"o",
|
| 317 |
+
"31",
|
| 318 |
+
" ",
|
| 319 |
+
"\uff0c",
|
| 320 |
+
" ",
|
| 321 |
+
"s",
|
| 322 |
+
"\u0268",
|
| 323 |
+
"t",
|
| 324 |
+
"5",
|
| 325 |
+
"p",
|
| 326 |
+
"a",
|
| 327 |
+
"u",
|
| 328 |
+
"31",
|
| 329 |
+
"m",
|
| 330 |
+
"a",
|
| 331 |
+
"\u014b",
|
| 332 |
+
"11",
|
| 333 |
+
" ",
|
| 334 |
+
"\uff0c",
|
| 335 |
+
" ",
|
| 336 |
+
"s",
|
| 337 |
+
"\u0268",
|
| 338 |
+
"t",
|
| 339 |
+
"5",
|
| 340 |
+
"p",
|
| 341 |
+
"a",
|
| 342 |
+
"u",
|
| 343 |
+
"31",
|
| 344 |
+
" ",
|
| 345 |
+
"l",
|
| 346 |
+
"e",
|
| 347 |
+
"24",
|
| 348 |
+
" ",
|
| 349 |
+
"\uff0c",
|
| 350 |
+
" ",
|
| 351 |
+
"n",
|
| 352 |
+
"\u0329",
|
| 353 |
+
"11",
|
| 354 |
+
" ",
|
| 355 |
+
"\u014b",
|
| 356 |
+
"a",
|
| 357 |
+
"i",
|
| 358 |
+
"11",
|
| 359 |
+
" ",
|
| 360 |
+
"k",
|
| 361 |
+
"i",
|
| 362 |
+
"11",
|
| 363 |
+
" ",
|
| 364 |
+
"\uff0c",
|
| 365 |
+
" ",
|
| 366 |
+
"\u014b",
|
| 367 |
+
"\u0329",
|
| 368 |
+
"11",
|
| 369 |
+
"k",
|
| 370 |
+
"o",
|
| 371 |
+
"\u014b",
|
| 372 |
+
"31",
|
| 373 |
+
"m",
|
| 374 |
+
"a",
|
| 375 |
+
"31",
|
| 376 |
+
"k",
|
| 377 |
+
"e",
|
| 378 |
+
"55",
|
| 379 |
+
" ",
|
| 380 |
+
"\uff0c",
|
| 381 |
+
" ",
|
| 382 |
+
"n",
|
| 383 |
+
"\u0329",
|
| 384 |
+
"11",
|
| 385 |
+
" ",
|
| 386 |
+
"h",
|
| 387 |
+
"e",
|
| 388 |
+
"55",
|
| 389 |
+
"m",
|
| 390 |
+
"o",
|
| 391 |
+
"11",
|
| 392 |
+
" ",
|
| 393 |
+
"s",
|
| 394 |
+
"a",
|
| 395 |
+
"11",
|
| 396 |
+
" ",
|
| 397 |
+
"\uff0c",
|
| 398 |
+
" ",
|
| 399 |
+
"\u014b",
|
| 400 |
+
"a",
|
| 401 |
+
"i",
|
| 402 |
+
"11",
|
| 403 |
+
" ",
|
| 404 |
+
"h",
|
| 405 |
+
"e",
|
| 406 |
+
"55",
|
| 407 |
+
" ",
|
| 408 |
+
"a",
|
| 409 |
+
"24",
|
| 410 |
+
" ",
|
| 411 |
+
"l",
|
| 412 |
+
"i",
|
| 413 |
+
"n",
|
| 414 |
+
"11",
|
| 415 |
+
" ",
|
| 416 |
+
"\uff0c",
|
| 417 |
+
" ",
|
| 418 |
+
"a",
|
| 419 |
+
"n",
|
| 420 |
+
"31",
|
| 421 |
+
"k",
|
| 422 |
+
"i",
|
| 423 |
+
"u",
|
| 424 |
+
"31",
|
| 425 |
+
"m",
|
| 426 |
+
"o",
|
| 427 |
+
"11",
|
| 428 |
+
"k",
|
| 429 |
+
"\u02b0",
|
| 430 |
+
"o",
|
| 431 |
+
"n",
|
| 432 |
+
"55",
|
| 433 |
+
"t",
|
| 434 |
+
"o",
|
| 435 |
+
"31",
|
| 436 |
+
" ",
|
| 437 |
+
"\uff0c",
|
| 438 |
+
" ",
|
| 439 |
+
"\u014b",
|
| 440 |
+
"i",
|
| 441 |
+
"11",
|
| 442 |
+
"h",
|
| 443 |
+
"o",
|
| 444 |
+
"31",
|
| 445 |
+
" ",
|
| 446 |
+
"m",
|
| 447 |
+
"o",
|
| 448 |
+
"11"
|
| 449 |
+
],
|
| 450 |
+
"XM",
|
| 451 |
+
null,
|
| 452 |
+
"sixian"
|
| 453 |
+
],
|
| 454 |
+
[
|
| 455 |
+
[
|
| 456 |
+
"t",
|
| 457 |
+
"\u02b0",
|
| 458 |
+
"e",
|
| 459 |
+
"u",
|
| 460 |
+
"55",
|
| 461 |
+
"t",
|
| 462 |
+
"\u0361",
|
| 463 |
+
"s",
|
| 464 |
+
"\u02b0",
|
| 465 |
+
"i",
|
| 466 |
+
"e",
|
| 467 |
+
"n",
|
| 468 |
+
"55",
|
| 469 |
+
" ",
|
| 470 |
+
"v",
|
| 471 |
+
"a",
|
| 472 |
+
"t",
|
| 473 |
+
"5",
|
| 474 |
+
" ",
|
| 475 |
+
"\u0288",
|
| 476 |
+
"\u0361",
|
| 477 |
+
"\u0282",
|
| 478 |
+
"o",
|
| 479 |
+
"n",
|
| 480 |
+
"24",
|
| 481 |
+
" ",
|
| 482 |
+
"\u0288",
|
| 483 |
+
"\u0361",
|
| 484 |
+
"\u0282",
|
| 485 |
+
"i",
|
| 486 |
+
"n",
|
| 487 |
+
"11",
|
| 488 |
+
"p",
|
| 489 |
+
"\u02b0",
|
| 490 |
+
"i",
|
| 491 |
+
"e",
|
| 492 |
+
"n",
|
| 493 |
+
"24",
|
| 494 |
+
" ",
|
| 495 |
+
"t",
|
| 496 |
+
"\u0361",
|
| 497 |
+
"s",
|
| 498 |
+
"\u02b0",
|
| 499 |
+
"i",
|
| 500 |
+
"u",
|
| 501 |
+
"55",
|
| 502 |
+
" ",
|
| 503 |
+
"t",
|
| 504 |
+
"o",
|
| 505 |
+
"11",
|
| 506 |
+
" ",
|
| 507 |
+
"t",
|
| 508 |
+
"\u02b0",
|
| 509 |
+
"o",
|
| 510 |
+
"i",
|
| 511 |
+
"55",
|
| 512 |
+
"p",
|
| 513 |
+
"e",
|
| 514 |
+
"t",
|
| 515 |
+
"5",
|
| 516 |
+
" ",
|
| 517 |
+
"\u0282",
|
| 518 |
+
"i",
|
| 519 |
+
"33",
|
| 520 |
+
" ",
|
| 521 |
+
"\u0288",
|
| 522 |
+
"\u0361",
|
| 523 |
+
"\u0282",
|
| 524 |
+
"i",
|
| 525 |
+
"n",
|
| 526 |
+
"11",
|
| 527 |
+
"f",
|
| 528 |
+
"u",
|
| 529 |
+
"24",
|
| 530 |
+
" ",
|
| 531 |
+
"l",
|
| 532 |
+
"e",
|
| 533 |
+
"24"
|
| 534 |
+
],
|
| 535 |
+
"HF",
|
| 536 |
+
null,
|
| 537 |
+
"hailu"
|
| 538 |
+
],
|
| 539 |
+
[
|
| 540 |
+
[
|
| 541 |
+
"t",
|
| 542 |
+
"\u02b0",
|
| 543 |
+
"e",
|
| 544 |
+
"u",
|
| 545 |
+
"55",
|
| 546 |
+
"t",
|
| 547 |
+
"\u0361",
|
| 548 |
+
"s",
|
| 549 |
+
"\u02b0",
|
| 550 |
+
"i",
|
| 551 |
+
"e",
|
| 552 |
+
"n",
|
| 553 |
+
"55",
|
| 554 |
+
" ",
|
| 555 |
+
"v",
|
| 556 |
+
"a",
|
| 557 |
+
"t",
|
| 558 |
+
"5",
|
| 559 |
+
" ",
|
| 560 |
+
"\u0288",
|
| 561 |
+
"\u0361",
|
| 562 |
+
"\u0282",
|
| 563 |
+
"o",
|
| 564 |
+
"n",
|
| 565 |
+
"24",
|
| 566 |
+
" ",
|
| 567 |
+
"\u0288",
|
| 568 |
+
"\u0361",
|
| 569 |
+
"\u0282",
|
| 570 |
+
"i",
|
| 571 |
+
"n",
|
| 572 |
+
"11",
|
| 573 |
+
"p",
|
| 574 |
+
"\u02b0",
|
| 575 |
+
"i",
|
| 576 |
+
"e",
|
| 577 |
+
"n",
|
| 578 |
+
"24",
|
| 579 |
+
" ",
|
| 580 |
+
"t",
|
| 581 |
+
"\u0361",
|
| 582 |
+
"s",
|
| 583 |
+
"\u02b0",
|
| 584 |
+
"i",
|
| 585 |
+
"u",
|
| 586 |
+
"55",
|
| 587 |
+
" ",
|
| 588 |
+
"t",
|
| 589 |
+
"o",
|
| 590 |
+
"11",
|
| 591 |
+
" ",
|
| 592 |
+
"t",
|
| 593 |
+
"\u02b0",
|
| 594 |
+
"o",
|
| 595 |
+
"i",
|
| 596 |
+
"55",
|
| 597 |
+
"p",
|
| 598 |
+
"e",
|
| 599 |
+
"t",
|
| 600 |
+
"5",
|
| 601 |
+
" ",
|
| 602 |
+
"\u0282",
|
| 603 |
+
"i",
|
| 604 |
+
"33",
|
| 605 |
+
" ",
|
| 606 |
+
"\u0288",
|
| 607 |
+
"\u0361",
|
| 608 |
+
"\u0282",
|
| 609 |
+
"i",
|
| 610 |
+
"n",
|
| 611 |
+
"11",
|
| 612 |
+
"f",
|
| 613 |
+
"u",
|
| 614 |
+
"24",
|
| 615 |
+
" ",
|
| 616 |
+
"l",
|
| 617 |
+
"e",
|
| 618 |
+
"24"
|
| 619 |
+
],
|
| 620 |
+
"HM",
|
| 621 |
+
null,
|
| 622 |
+
"hailu"
|
| 623 |
+
]
|
| 624 |
+
],
|
| 625 |
+
"eval_split_max_size": null,
|
| 626 |
+
"eval_split_size": 0.01,
|
| 627 |
+
"use_speaker_weighted_sampler": false,
|
| 628 |
+
"speaker_weighted_sampler_alpha": 1.0,
|
| 629 |
+
"use_language_weighted_sampler": false,
|
| 630 |
+
"language_weighted_sampler_alpha": 1.0,
|
| 631 |
+
"use_length_weighted_sampler": false,
|
| 632 |
+
"length_weighted_sampler_alpha": 1.0,
|
| 633 |
+
"model_args": {
|
| 634 |
+
"num_chars": 38,
|
| 635 |
+
"out_channels": 513,
|
| 636 |
+
"spec_segment_size": 32,
|
| 637 |
+
"hidden_channels": 192,
|
| 638 |
+
"hidden_channels_ffn_text_encoder": 768,
|
| 639 |
+
"num_heads_text_encoder": 2,
|
| 640 |
+
"num_layers_text_encoder": 6,
|
| 641 |
+
"kernel_size_text_encoder": 3,
|
| 642 |
+
"dropout_p_text_encoder": 0.1,
|
| 643 |
+
"dropout_p_duration_predictor": 0.5,
|
| 644 |
+
"kernel_size_posterior_encoder": 5,
|
| 645 |
+
"dilation_rate_posterior_encoder": 1,
|
| 646 |
+
"num_layers_posterior_encoder": 16,
|
| 647 |
+
"kernel_size_flow": 5,
|
| 648 |
+
"dilation_rate_flow": 1,
|
| 649 |
+
"num_layers_flow": 4,
|
| 650 |
+
"resblock_type_decoder": "1",
|
| 651 |
+
"resblock_kernel_sizes_decoder": [
|
| 652 |
+
3,
|
| 653 |
+
7,
|
| 654 |
+
11
|
| 655 |
+
],
|
| 656 |
+
"resblock_dilation_sizes_decoder": [
|
| 657 |
+
[
|
| 658 |
+
1,
|
| 659 |
+
3,
|
| 660 |
+
5
|
| 661 |
+
],
|
| 662 |
+
[
|
| 663 |
+
1,
|
| 664 |
+
3,
|
| 665 |
+
5
|
| 666 |
+
],
|
| 667 |
+
[
|
| 668 |
+
1,
|
| 669 |
+
3,
|
| 670 |
+
5
|
| 671 |
+
]
|
| 672 |
+
],
|
| 673 |
+
"upsample_rates_decoder": [
|
| 674 |
+
8,
|
| 675 |
+
8,
|
| 676 |
+
2,
|
| 677 |
+
2
|
| 678 |
+
],
|
| 679 |
+
"upsample_initial_channel_decoder": 512,
|
| 680 |
+
"upsample_kernel_sizes_decoder": [
|
| 681 |
+
16,
|
| 682 |
+
16,
|
| 683 |
+
4,
|
| 684 |
+
4
|
| 685 |
+
],
|
| 686 |
+
"periods_multi_period_discriminator": [
|
| 687 |
+
2,
|
| 688 |
+
3,
|
| 689 |
+
5,
|
| 690 |
+
7,
|
| 691 |
+
11
|
| 692 |
+
],
|
| 693 |
+
"use_sdp": true,
|
| 694 |
+
"noise_scale": 1.0,
|
| 695 |
+
"inference_noise_scale": 0.667,
|
| 696 |
+
"length_scale": 1,
|
| 697 |
+
"noise_scale_dp": 1.0,
|
| 698 |
+
"inference_noise_scale_dp": 1.0,
|
| 699 |
+
"max_inference_len": null,
|
| 700 |
+
"init_discriminator": true,
|
| 701 |
+
"use_spectral_norm_disriminator": false,
|
| 702 |
+
"use_speaker_embedding": true,
|
| 703 |
+
"num_speakers": 4,
|
| 704 |
+
"speakers_file": "speakers.pth",
|
| 705 |
+
"d_vector_file": null,
|
| 706 |
+
"speaker_embedding_channels": 256,
|
| 707 |
+
"use_d_vector_file": false,
|
| 708 |
+
"d_vector_dim": 0,
|
| 709 |
+
"detach_dp_input": true,
|
| 710 |
+
"use_language_embedding": true,
|
| 711 |
+
"embedded_language_dim": 4,
|
| 712 |
+
"num_languages": 2,
|
| 713 |
+
"language_ids_file": "language_ids.json",
|
| 714 |
+
"use_speaker_encoder_as_loss": false,
|
| 715 |
+
"speaker_encoder_config_path": "",
|
| 716 |
+
"speaker_encoder_model_path": "",
|
| 717 |
+
"condition_dp_on_speaker": true,
|
| 718 |
+
"freeze_encoder": false,
|
| 719 |
+
"freeze_DP": false,
|
| 720 |
+
"freeze_PE": false,
|
| 721 |
+
"freeze_flow_decoder": false,
|
| 722 |
+
"freeze_waveform_decoder": false,
|
| 723 |
+
"encoder_sample_rate": null,
|
| 724 |
+
"interpolate_z": true,
|
| 725 |
+
"reinit_DP": false,
|
| 726 |
+
"reinit_text_encoder": false
|
| 727 |
+
},
|
| 728 |
+
"lr_gen": 0.0002,
|
| 729 |
+
"lr_disc": 0.0002,
|
| 730 |
+
"lr_scheduler_gen": "ExponentialLR",
|
| 731 |
+
"lr_scheduler_gen_params": {
|
| 732 |
+
"gamma": 0.999875,
|
| 733 |
+
"last_epoch": -1
|
| 734 |
+
},
|
| 735 |
+
"lr_scheduler_disc": "ExponentialLR",
|
| 736 |
+
"lr_scheduler_disc_params": {
|
| 737 |
+
"gamma": 0.999875,
|
| 738 |
+
"last_epoch": -1
|
| 739 |
+
},
|
| 740 |
+
"kl_loss_alpha": 1.0,
|
| 741 |
+
"disc_loss_alpha": 1.0,
|
| 742 |
+
"gen_loss_alpha": 1.0,
|
| 743 |
+
"feat_loss_alpha": 1.0,
|
| 744 |
+
"mel_loss_alpha": 45.0,
|
| 745 |
+
"dur_loss_alpha": 1.0,
|
| 746 |
+
"speaker_encoder_loss_alpha": 1.0,
|
| 747 |
+
"return_wav": true,
|
| 748 |
+
"use_weighted_sampler": false,
|
| 749 |
+
"weighted_sampler_attrs": {},
|
| 750 |
+
"weighted_sampler_multipliers": {},
|
| 751 |
+
"r": 1,
|
| 752 |
+
"num_speakers": 0,
|
| 753 |
+
"use_speaker_embedding": true,
|
| 754 |
+
"speakers_file": "speakers.pth",
|
| 755 |
+
"speaker_embedding_channels": 256,
|
| 756 |
+
"language_ids_file": "language_ids.json",
|
| 757 |
+
"use_language_embedding": true,
|
| 758 |
+
"use_d_vector_file": false,
|
| 759 |
+
"d_vector_file": null,
|
| 760 |
+
"d_vector_dim": 0
|
| 761 |
+
}
|
language_ids.json
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"hailu": 0,
|
| 3 |
+
"sixian": 1
|
| 4 |
+
}
|
model.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b7a8ac78e8b73e97ac3d47a33b425a7d96fa23504ee90529fce5ca164f8a97b1
|
| 3 |
+
size 1039828314
|
speakers.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:90ee24767f2f887e7228d9241d390d4739d5268e5ecfb0f473f6397860bea37c
|
| 3 |
+
size 864
|