Hugging Face's logo

Yehor
/
radtts-uk

Text-to-Speech
TensorBoard
Ukrainian
Model card Files
xet
 Metrics Community
Yehor commited on
Commit
6bb6e4b
·
verified ·
1 Parent(s): f3402b8

Upload 5 files

Browse files
Files changed (5) hide show
  1. radtts-pp-dap-model/config.json +218 -0
  2. radtts-pp-dap-model/logs/events.out.tfevents.1683879874.mllab-MS-7B09.23965.0 +3 -0
  3. radtts-pp-dap-model/logs/events.out.tfevents.1683879936.mllab-MS-7B09.24044.0 +3 -0
  4. radtts-pp-dap-model/logs/events.out.tfevents.1683880022.mllab-MS-7B09.24149.0 +3 -0
  5. radtts-pp-dap-model/model_dap_84000.pt +3 -0
radtts-pp-dap-model/config.json ADDED
@@ -0,0 +1,218 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train_config": {
3
+ "output_directory": "outdir_pp_model",
4
+ "epochs": 10000000,
5
+ "optim_algo": "RAdam",
6
+ "learning_rate": 0.001,
7
+ "weight_decay": 1e-06,
8
+ "sigma": 1.0,
9
+ "iters_per_checkpoint": 1000,
10
+ "batch_size": 16,
11
+ "seed": null,
12
+ "checkpoint_path": "",
13
+ "ignore_layers": [],
14
+ "ignore_layers_warmstart": [],
15
+ "finetune_layers": [],
16
+ "include_layers": [],
17
+ "vocoder_config_path": "models/hifigan_22khz_config.json",
18
+ "vocoder_checkpoint_path": "models/hifigan_ljs_generator_v1.pt",
19
+ "log_attribute_samples": true,
20
+ "log_decoder_samples": true,
21
+ "warmstart_checkpoint_path": "outdir_pp/model_100000",
22
+ "use_amp": true,
23
+ "grad_clip_val": 1.0,
24
+ "loss_weights": {
25
+ "blank_logprob": -1,
26
+ "ctc_loss_weight": 0.1,
27
+ "binarization_loss_weight": 1.0,
28
+ "dur_loss_weight": 1.0,
29
+ "f0_loss_weight": 1.0,
30
+ "energy_loss_weight": 1.0,
31
+ "vpred_loss_weight": 1.0
32
+ },
33
+ "binarization_start_iter": 0,
34
+ "kl_loss_start_iter": 0,
35
+ "unfreeze_modules": "all"
36
+ },
37
+ "data_config": {
38
+ "training_files": {
39
+ "LJS": {
40
+ "basedir": "filelists/",
41
+ "audiodir": "wavs",
42
+ "filelist": "3speakers_ukrainian_train_filelist_dc.txt",
43
+ "lmdbpath": ""
44
+ }
45
+ },
46
+ "validation_files": {
47
+ "LJS": {
48
+ "basedir": "filelists/",
49
+ "audiodir": "wavs",
50
+ "filelist": "3speakers_ukrainian_val_filelist_dc.txt",
51
+ "lmdbpath": ""
52
+ }
53
+ },
54
+ "dur_min": 0.1,
55
+ "dur_max": 10.2,
56
+ "sampling_rate": 22050,
57
+ "filter_length": 1024,
58
+ "hop_length": 256,
59
+ "win_length": 1024,
60
+ "n_mel_channels": 80,
61
+ "mel_fmin": 0.0,
62
+ "mel_fmax": 8000.0,
63
+ "f0_min": 80.0,
64
+ "f0_max": 640.0,
65
+ "max_wav_value": 32768.0,
66
+ "use_f0": true,
67
+ "use_log_f0": 0,
68
+ "use_energy_avg": true,
69
+ "use_scaled_energy": true,
70
+ "symbol_set": "ukrainian",
71
+ "cleaner_names": [
72
+ "ukrainian_cleaners"
73
+ ],
74
+ "heteronyms_path": "tts_text_processing/heteronyms",
75
+ "phoneme_dict_path": "tts_text_processing/cmudict-0.7b",
76
+ "p_phoneme": 0.0,
77
+ "handle_phoneme": "word",
78
+ "handle_phoneme_ambiguous": "ignore",
79
+ "include_speakers": null,
80
+ "n_frames": -1,
81
+ "betabinom_cache_path": "/home/dmytro_chaplinsky/RAD-TTS/radtts-code/cache",
82
+ "lmdb_cache_path": "",
83
+ "use_attn_prior_masking": true,
84
+ "prepend_space_to_text": true,
85
+ "append_space_to_text": true,
86
+ "add_bos_eos_to_text": false,
87
+ "betabinom_scaling_factor": 1.0,
88
+ "distance_tx_unvoiced": false,
89
+ "mel_noise_scale": 0.0
90
+ },
91
+ "dist_config": {
92
+ "dist_backend": "nccl",
93
+ "dist_url": "tcp://localhost:54321"
94
+ },
95
+ "model_config": {
96
+ "n_speakers": 3,
97
+ "n_speaker_dim": 16,
98
+ "n_text": 185,
99
+ "n_text_dim": 512,
100
+ "n_flows": 8,
101
+ "n_conv_layers_per_step": 4,
102
+ "n_mel_channels": 80,
103
+ "n_hidden": 1024,
104
+ "mel_encoder_n_hidden": 512,
105
+ "dummy_speaker_embedding": false,
106
+ "n_early_size": 2,
107
+ "n_early_every": 2,
108
+ "n_group_size": 2,
109
+ "affine_model": "wavenet",
110
+ "include_modules": "decatndpmvpredapm",
111
+ "scaling_fn": "tanh",
112
+ "matrix_decomposition": "LUS",
113
+ "learn_alignments": true,
114
+ "use_speaker_emb_for_alignment": false,
115
+ "attn_straight_through_estimator": true,
116
+ "use_context_lstm": true,
117
+ "context_lstm_norm": "spectral",
118
+ "context_lstm_w_f0_and_energy": true,
119
+ "text_encoder_lstm_norm": "spectral",
120
+ "n_f0_dims": 1,
121
+ "n_energy_avg_dims": 1,
122
+ "use_first_order_features": false,
123
+ "unvoiced_bias_activation": "relu",
124
+ "decoder_use_partial_padding": true,
125
+ "decoder_use_unvoiced_bias": true,
126
+ "ap_pred_log_f0": true,
127
+ "ap_use_unvoiced_bias": false,
128
+ "ap_use_voiced_embeddings": true,
129
+ "dur_model_config": {
130
+ "name": "dap",
131
+ "hparams": {
132
+ "n_speaker_dim": 16,
133
+ "bottleneck_hparams": {
134
+ "in_dim": 512,
135
+ "reduction_factor": 16,
136
+ "norm": "weightnorm",
137
+ "non_linearity": "relu"
138
+ },
139
+ "take_log_of_input": true,
140
+ "arch_hparams": {
141
+ "out_dim": 1,
142
+ "n_layers": 2,
143
+ "n_channels": 256,
144
+ "kernel_size": 3,
145
+ "p_dropout": 0.25,
146
+ "in_dim": 48
147
+ }
148
+ }
149
+ },
150
+ "f0_model_config": {
151
+ "name": "dap",
152
+ "hparams": {
153
+ "n_speaker_dim": 16,
154
+ "bottleneck_hparams": {
155
+ "in_dim": 512,
156
+ "reduction_factor": 16,
157
+ "norm": "weightnorm",
158
+ "non_linearity": "relu"
159
+ },
160
+ "take_log_of_input": false,
161
+ "use_transformer": false,
162
+ "arch_hparams": {
163
+ "out_dim": 1,
164
+ "n_layers": 2,
165
+ "n_channels": 256,
166
+ "kernel_size": 11,
167
+ "p_dropout": 0.5,
168
+ "in_dim": 48
169
+ }
170
+ }
171
+ },
172
+ "energy_model_config": {
173
+ "name": "dap",
174
+ "hparams": {
175
+ "n_speaker_dim": 16,
176
+ "bottleneck_hparams": {
177
+ "in_dim": 512,
178
+ "reduction_factor": 16,
179
+ "norm": "weightnorm",
180
+ "non_linearity": "relu"
181
+ },
182
+ "take_log_of_input": false,
183
+ "use_transformer": false,
184
+ "arch_hparams": {
185
+ "out_dim": 1,
186
+ "n_layers": 2,
187
+ "n_channels": 256,
188
+ "kernel_size": 3,
189
+ "p_dropout": 0.25,
190
+ "in_dim": 48
191
+ }
192
+ }
193
+ },
194
+ "v_model_config": {
195
+ "name": "dap",
196
+ "hparams": {
197
+ "n_speaker_dim": 16,
198
+ "take_log_of_input": false,
199
+ "bottleneck_hparams": {
200
+ "in_dim": 512,
201
+ "reduction_factor": 16,
202
+ "norm": "weightnorm",
203
+ "non_linearity": "relu"
204
+ },
205
+ "arch_hparams": {
206
+ "out_dim": 1,
207
+ "n_layers": 2,
208
+ "n_channels": 256,
209
+ "kernel_size": 3,
210
+ "p_dropout": 0.5,
211
+ "lstm_type": "",
212
+ "use_linear": 1,
213
+ "in_dim": 48
214
+ }
215
+ }
216
+ }
217
+ }
218
+ }
radtts-pp-dap-model/logs/events.out.tfevents.1683879874.mllab-MS-7B09.23965.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96f132b16fca59671d9bdee5a5b7e0b1f8efe93f49802fecaed06a16a770ed42
3
+ size 88
radtts-pp-dap-model/logs/events.out.tfevents.1683879936.mllab-MS-7B09.24044.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e08c7277a5b5e4202f52e34dcaa798d467da9abda601297998eeba68c6cfb44e
3
+ size 512
radtts-pp-dap-model/logs/events.out.tfevents.1683880022.mllab-MS-7B09.24149.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eade992e74e6d86fd44a38037308cb344f229c80703aa0dc5e7aac28eaf6ab3b
3
+ size 138829675
radtts-pp-dap-model/model_dap_84000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:460d1e09a1116d833d8d0532dc9dfd417c15eb1c675108c64dfaf9ebe37f5941
3
+ size 2762780713