alp commited on
Commit
3922ca3
verified
1 Parent(s): 250848e

Upload 2 files

Browse files
fast-speech_best_model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a5aefb9f49f6172e34b816e1de8f5234012f0a9a05747973f6610e40869983f
3
+ size 457921637
fast-speech_config.json ADDED
@@ -0,0 +1,213 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "output_path": "/home/twbgmy/play/TTS-play/TTS/recipes/catotron",
3
+ "logger_uri": null,
4
+ "run_name": "fast_pitch_ljspeech",
5
+ "project_name": null,
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": "tensorboard",
12
+ "log_model_step": null,
13
+ "save_step": 10000,
14
+ "save_n_checkpoints": 5,
15
+ "save_checkpoints": true,
16
+ "save_all_best": false,
17
+ "save_best_after": 1000,
18
+ "target_loss": null,
19
+ "print_eval": false,
20
+ "test_delay_epochs": -1,
21
+ "run_eval": true,
22
+ "run_eval_steps": null,
23
+ "distributed_backend": "nccl",
24
+ "distributed_url": "tcp://localhost:54321",
25
+ "mixed_precision": false,
26
+ "epochs": 1000,
27
+ "batch_size": 16,
28
+ "eval_batch_size": 16,
29
+ "grad_clip": 5.0,
30
+ "scheduler_after_epoch": true,
31
+ "lr": 0.0001,
32
+ "optimizer": "Adam",
33
+ "optimizer_params": {
34
+ "betas": [
35
+ 0.9,
36
+ 0.998
37
+ ],
38
+ "weight_decay": 1e-06
39
+ },
40
+ "lr_scheduler": "NoamLR",
41
+ "lr_scheduler_params": {
42
+ "warmup_steps": 4000
43
+ },
44
+ "use_grad_scaler": false,
45
+ "cudnn_enable": true,
46
+ "cudnn_deterministic": false,
47
+ "cudnn_benchmark": false,
48
+ "training_seed": 54321,
49
+ "model": "fast_pitch",
50
+ "num_loader_workers": 8,
51
+ "num_eval_loader_workers": 4,
52
+ "use_noise_augment": false,
53
+ "audio": {
54
+ "fft_size": 1024,
55
+ "win_length": 1024,
56
+ "hop_length": 256,
57
+ "frame_shift_ms": null,
58
+ "frame_length_ms": null,
59
+ "stft_pad_mode": "reflect",
60
+ "sample_rate": 22050,
61
+ "resample": false,
62
+ "preemphasis": 0.0,
63
+ "ref_level_db": 20,
64
+ "do_sound_norm": false,
65
+ "log_func": "np.log",
66
+ "do_trim_silence": true,
67
+ "trim_db": 60.0,
68
+ "do_rms_norm": false,
69
+ "db_level": null,
70
+ "power": 1.5,
71
+ "griffin_lim_iters": 60,
72
+ "num_mels": 80,
73
+ "mel_fmin": 0.0,
74
+ "mel_fmax": 8000,
75
+ "spec_gain": 1.0,
76
+ "do_amp_to_db_linear": true,
77
+ "do_amp_to_db_mel": true,
78
+ "pitch_fmax": 640.0,
79
+ "pitch_fmin": 0.0,
80
+ "signal_norm": false,
81
+ "min_level_db": -100,
82
+ "symmetric_norm": true,
83
+ "max_norm": 4.0,
84
+ "clip_norm": true,
85
+ "stats_path": null
86
+ },
87
+ "use_phonemes": false,
88
+ "phonemizer": null,
89
+ "phoneme_language": "ca-es",
90
+ "compute_input_seq_cache": true,
91
+ "text_cleaner": "multilingual_cleaners",
92
+ "enable_eos_bos_chars": false,
93
+ "test_sentences_file": "",
94
+ "phoneme_cache_path": null,
95
+ "characters": {
96
+ "characters_class": "TTS.tts.utils.text.characters.Graphemes",
97
+ "vocab_dict": null,
98
+ "pad": "_",
99
+ "eos": "*",
100
+ "bos": "^",
101
+ "blank": null,
102
+ "characters": "A\u00c0\u00c1BC\u00c7DE\u00c9\u00c8FGHI\u00cd\u00cfJKLMNO\u00d3\u00d2PQRSTU\u00dc\u00daVWXYZa\u00e0\u00e1bc\u00e7de\u00e9\u00e8fghi\u00ed\u00efjklmno\u00f3\u00f2pqrstu\u00fc\u00favwxyz",
103
+ "punctuations": "!'(),-.:;?\u00b7 ",
104
+ "phonemes": "",
105
+ "is_unique": true,
106
+ "is_sorted": true
107
+ },
108
+ "add_blank": false,
109
+ "batch_group_size": 0,
110
+ "loss_masking": null,
111
+ "min_audio_len": 1,
112
+ "max_audio_len": Infinity,
113
+ "min_text_len": 1,
114
+ "max_text_len": Infinity,
115
+ "compute_f0": true,
116
+ "compute_linear_spec": false,
117
+ "precompute_num_workers": 4,
118
+ "start_by_longest": false,
119
+ "datasets": [
120
+ {
121
+ "name": "custom_turkish",
122
+ "path": "/home/twbgmy/play/TTS-play/TTS/recipes/catotron/upc_ona",
123
+ "meta_file_train": "upc_ona_train.txt",
124
+ "ignored_speakers": null,
125
+ "language": "",
126
+ "meta_file_val": "",
127
+ "meta_file_attn_mask": ""
128
+ },
129
+ {
130
+ "name": "custom_turkish",
131
+ "path": "/home/twbgmy/play/TTS-play/TTS/recipes/catotron/upc_ona",
132
+ "meta_file_train": "upc_ona_val.txt",
133
+ "ignored_speakers": null,
134
+ "language": "",
135
+ "meta_file_val": "",
136
+ "meta_file_attn_mask": ""
137
+ }
138
+ ],
139
+ "test_sentences": [
140
+ "Hola Barcelona!",
141
+ "Escriviu al text."
142
+ ],
143
+ "eval_split_max_size": null,
144
+ "eval_split_size": 0.01,
145
+ "use_speaker_weighted_sampler": false,
146
+ "speaker_weighted_sampler_alpha": 1.0,
147
+ "use_language_weighted_sampler": false,
148
+ "language_weighted_sampler_alpha": 1.0,
149
+ "use_length_weighted_sampler": false,
150
+ "length_weighted_sampler_alpha": 1.0,
151
+ "base_model": "forward_tts",
152
+ "model_args": {
153
+ "num_chars": 89,
154
+ "out_channels": 80,
155
+ "hidden_channels": 384,
156
+ "use_aligner": true,
157
+ "use_pitch": true,
158
+ "pitch_predictor_hidden_channels": 256,
159
+ "pitch_predictor_kernel_size": 3,
160
+ "pitch_predictor_dropout_p": 0.1,
161
+ "pitch_embedding_kernel_size": 3,
162
+ "duration_predictor_hidden_channels": 256,
163
+ "duration_predictor_kernel_size": 3,
164
+ "duration_predictor_dropout_p": 0.1,
165
+ "positional_encoding": true,
166
+ "poisitonal_encoding_use_scale": true,
167
+ "length_scale": 1,
168
+ "encoder_type": "fftransformer",
169
+ "encoder_params": {
170
+ "hidden_channels_ffn": 1024,
171
+ "num_heads": 1,
172
+ "num_layers": 6,
173
+ "dropout_p": 0.1
174
+ },
175
+ "decoder_type": "fftransformer",
176
+ "decoder_params": {
177
+ "hidden_channels_ffn": 1024,
178
+ "num_heads": 1,
179
+ "num_layers": 6,
180
+ "dropout_p": 0.1
181
+ },
182
+ "detach_duration_predictor": false,
183
+ "max_duration": 75,
184
+ "num_speakers": 1,
185
+ "use_speaker_embedding": false,
186
+ "speakers_file": null,
187
+ "use_d_vector_file": false,
188
+ "d_vector_dim": null,
189
+ "d_vector_file": null
190
+ },
191
+ "num_speakers": 0,
192
+ "speakers_file": null,
193
+ "use_speaker_embedding": false,
194
+ "use_d_vector_file": false,
195
+ "d_vector_file": false,
196
+ "d_vector_dim": 0,
197
+ "spec_loss_type": "mse",
198
+ "duration_loss_type": "mse",
199
+ "use_ssim_loss": true,
200
+ "ssim_loss_alpha": 1.0,
201
+ "spec_loss_alpha": 1.0,
202
+ "aligner_loss_alpha": 1.0,
203
+ "pitch_loss_alpha": 0.1,
204
+ "dur_loss_alpha": 0.1,
205
+ "binary_align_loss_alpha": 0.1,
206
+ "binary_loss_warmup_epochs": 150,
207
+ "min_seq_len": 13,
208
+ "max_seq_len": 500000,
209
+ "r": 1,
210
+ "f0_cache_path": "/home/twbgmy/play/TTS-play/TTS/recipes/catotron/f0_cache",
211
+ "restore_path": "/home/twbgmy/.local/share/tts/tts_models--en--ljspeech--fast_pitch/model_file.pth",
212
+ "github_branch": "* dev"
213
+ }