ZLSCompLing/CoquiTTS-Max
099
1{
2 "output_path": "D:\\NEWTTS\\TTS\\dataset\\",
3 "logger_uri": null,
4 "run_name": "lod-male-vits-lb",
5 "project_name": null,
6 "run_description": "\ud83d\udc38Coqui trainer run.",
7 "print_step": 30,
8 "plot_step": 100,
9 "model_param_stats": false,
10 "wandb_entity": null,
11 "dashboard_logger": "tensorboard",
12 "save_on_interrupt": true,
13 "log_model_step": null,
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": 16,
31 "grad_clip": [
32 1000.0,
33 1000.0
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": 0,
56 "num_eval_loader_workers": 4,
57 "use_noise_augment": false,
58 "audio": {
59 "fft_size": 1024,
60 "sample_rate": 22050,
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": "phoneme_cleaners",
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": "abcdefghijklmnopqrstuvwxyz\u00b5\u00df\u00e0\u00e1\u00e2\u00e4\u00e5\u00e6\u00e7\u00e8\u00e9\u00ea\u00eb\u00ec\u00ed\u00ee\u00ef\u00f1\u00f2\u00f3\u00f4\u00f6\u00f9\u00fa\u00fb\u00fc\u0105\u0107\u0119\u0142\u0144\u0153\u015b\u015f\u017a\u017c\u0192\u0430\u0431\u0432\u0433\u0434\u0435\u0436\u0437\u0438\u0439\u043a\u043b\u043c\u043d\u043e\u043f\u0440\u0441\u0442\u0443\u0444\u0445\u0446\u0447\u0448\u0449\u044a\u044b\u044c\u044d\u044e\u044f\u0451\u0454\u0456\u0457\u0491\u04e7\u00ab\u00b0\u00b1\u00b5\u00bb$%&\u2018\u2019\u201a\u201c`\u201d\u201e\u00f5\u00e3\u0131",
83 "punctuations": "!\u00a1'(),-.:;\u00bf? ",
84 "phonemes": null,
85 "is_unique": true,
86 "is_sorted": true
87 },
88 "add_blank": true,
89 "batch_group_size": 5,
90 "loss_masking": null,
91 "min_audio_len": 1,
92 "max_audio_len": null,
93 "min_text_len": 1,
94 "max_text_len": null,
95 "compute_f0": false,
96 "compute_energy": false,
97 "compute_linear_spec": true,
98 "precompute_num_workers": 0,
99 "start_by_longest": false,
100 "shuffle": false,
101 "drop_last": false,
102 "datasets": [
103 {
104 "formatter": "luxembourgish",
105 "dataset_name": "",
106 "path": "D:\\NEWTTS\\TTS\\dataset\\lod-male-vits-lb/",
107 "meta_file_train": "metadata.csv",
108 "ignored_speakers": null,
109 "language": "",
110 "phonemizer": "",
111 "meta_file_val": "",
112 "meta_file_attn_mask": ""
113 }
114 ],
115 "test_sentences": [
116 "an der z\u00e4it hunn sech den nordwand an d\u2019sonn gestridden, wie vun hinnen zwee wuel m\u00e9i staark wier,",
117 "w\u00e9i e wanderer, deen an ee waarme mantel agepak war, iwwert de wee koum.",
118 "si goufen sech eens, datt deej\u00e9inege fir dee st\u00e4erkste g\u00eblle sollt, deen de wanderer forc\u00e9iere g\u00e9if, s\u00e4i mantel auszedoen.",
119 "dunn huet d\u2019sonn d\u2019loft mat hire fr\u00ebndleche strale gewiermt, a schonn no kuerzer z\u00e4it huet de wanderer s\u00e4i mantel ausgedoen.",
120 "do huet den rdwand missen zouginn, datt d\u2019sonn vun hinnen zwee dee st\u00e4erkste wier."
121 ],
122 "eval_split_max_size": null,
123 "eval_split_size": 0.01,
124 "use_speaker_weighted_sampler": false,
125 "speaker_weighted_sampler_alpha": 1.0,
126 "use_language_weighted_sampler": false,
127 "language_weighted_sampler_alpha": 1.0,
128 "use_length_weighted_sampler": false,
129 "length_weighted_sampler_alpha": 1.0,
130 "model_args": {
131 "num_chars": 134,
132 "out_channels": 513,
133 "spec_segment_size": 32,
134 "hidden_channels": 192,
135 "hidden_channels_ffn_text_encoder": 768,
136 "num_heads_text_encoder": 2,
137 "num_layers_text_encoder": 6,
138 "kernel_size_text_encoder": 3,
139 "dropout_p_text_encoder": 0.1,
140 "dropout_p_duration_predictor": 0.5,
141 "kernel_size_posterior_encoder": 5,
142 "dilation_rate_posterior_encoder": 1,
143 "num_layers_posterior_encoder": 16,
144 "kernel_size_flow": 5,
145 "dilation_rate_flow": 1,
146 "num_layers_flow": 4,
147 "resblock_type_decoder": "1",
148 "resblock_kernel_sizes_decoder": [
149 3,
150 7,
151 11
152 ],
153 "resblock_dilation_sizes_decoder": [
154 [
155 1,
156 3,
157 5
158 ],
159 [
160 1,
161 3,
162 5
163 ],
164 [
165 1,
166 3,
167 5
168 ]
169 ],
170 "upsample_rates_decoder": [
171 8,
172 8,
173 2,
174 2
175 ],
176 "upsample_initial_channel_decoder": 512,
177 "upsample_kernel_sizes_decoder": [
178 16,
179 16,
180 4,
181 4
182 ],
183 "periods_multi_period_discriminator": [
184 2,
185 3,
186 5,
187 7,
188 11
189 ],
190 "use_sdp": true,
191 "noise_scale": 1.0,
192 "inference_noise_scale": 0.667,
193 "length_scale": 1.0,
194 "noise_scale_dp": 1.0,
195 "inference_noise_scale_dp": 1.0,
196 "max_inference_len": null,
197 "init_discriminator": true,
198 "use_spectral_norm_disriminator": false,
199 "use_speaker_embedding": false,
200 "num_speakers": 0,
201 "speakers_file": null,
202 "d_vector_file": null,
203 "speaker_embedding_channels": 256,
204 "use_d_vector_file": false,
205 "d_vector_dim": 0,
206 "detach_dp_input": true,
207 "use_language_embedding": false,
208 "embedded_language_dim": 4,
209 "num_languages": 0,
210 "language_ids_file": null,
211 "use_speaker_encoder_as_loss": false,
212 "speaker_encoder_config_path": "",
213 "speaker_encoder_model_path": "",
214 "condition_dp_on_speaker": true,
215 "freeze_encoder": false,
216 "freeze_DP": false,
217 "freeze_PE": false,
218 "freeze_flow_decoder": false,
219 "freeze_waveform_decoder": false,
220 "encoder_sample_rate": null,
221 "interpolate_z": true,
222 "reinit_DP": false,
223 "reinit_text_encoder": false
224 },
225 "lr_gen": 0.0002,
226 "lr_disc": 0.0002,
227 "lr_scheduler_gen": "ExponentialLR",
228 "lr_scheduler_gen_params": {
229 "gamma": 0.999875,
230 "last_epoch": -1
231 },
232 "lr_scheduler_disc": "ExponentialLR",
233 "lr_scheduler_disc_params": {
234 "gamma": 0.999875,
235 "last_epoch": -1
236 },
237 "kl_loss_alpha": 1.0,
238 "disc_loss_alpha": 1.0,
239 "gen_loss_alpha": 1.0,
240 "feat_loss_alpha": 1.0,
241 "mel_loss_alpha": 45.0,
242 "dur_loss_alpha": 1.0,
243 "speaker_encoder_loss_alpha": 1.0,
244 "return_wav": true,
245 "use_weighted_sampler": false,
246 "weighted_sampler_attrs": {},
247 "weighted_sampler_multipliers": {},
248 "r": 1,
249 "num_speakers": 0,
250 "use_speaker_embedding": false,
251 "speakers_file": null,
252 "speaker_embedding_channels": 256,
253 "language_ids_file": null,
254 "use_language_embedding": false,
255 "use_d_vector_file": false,
256 "d_vector_file": null,
257 "d_vector_dim": 0,
258 "restore_path": "D:/NEWTTS/TTS/dataset/lod-male-vits-lb-June-28-2024_03+19PM-dbf1a08a/checkpoint_50000.pth",
259 "github_branch": "* dev"
260}