909ahmed commited on
Commit
d00e9b6
·
verified ·
1 Parent(s): c49ad7d

Upload 3 files

Browse files
Files changed (3) hide show
  1. best_model.pth +3 -0
  2. config.json +211 -0
  3. speakers.pth +3 -0
best_model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e785c1e6841d03854fdae311bac8c1338d212d36977549e3c2e852339bb5323
3
+ size 637339757
config.json ADDED
@@ -0,0 +1,211 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "output_path": "output_indic_fastpitch/tef13x",
3
+ "logger_uri": null,
4
+ "run_name": "tef13x_fastpitch_indictts_all_align_off",
5
+ "project_name": "indic-fastpitch-stage2",
6
+ "run_description": "align_off",
7
+ "print_step": 100,
8
+ "plot_step": 100,
9
+ "model_param_stats": false,
10
+ "wandb_entity": "indic-asr",
11
+ "dashboard_logger": "wandb",
12
+ "log_model_step": 10000,
13
+ "save_step": 10000,
14
+ "save_n_checkpoints": 1,
15
+ "save_checkpoints": true,
16
+ "save_all_best": false,
17
+ "save_best_after": 10000,
18
+ "target_loss": null,
19
+ "print_eval": false,
20
+ "test_delay_epochs": 0,
21
+ "run_eval": true,
22
+ "distributed_backend": "nccl",
23
+ "distributed_url": "tcp://localhost:54321",
24
+ "mixed_precision": true,
25
+ "epochs": 1000,
26
+ "batch_size": 32,
27
+ "eval_batch_size": 32,
28
+ "grad_clip": 5.0,
29
+ "scheduler_after_epoch": true,
30
+ "lr": 0.0001,
31
+ "optimizer": "Adam",
32
+ "optimizer_params": {
33
+ "betas": [
34
+ 0.9,
35
+ 0.998
36
+ ],
37
+ "weight_decay": 1e-06
38
+ },
39
+ "lr_scheduler": "NoamLR",
40
+ "lr_scheduler_params": {
41
+ "warmup_steps": 4000
42
+ },
43
+ "use_grad_scaler": false,
44
+ "cudnn_enable": true,
45
+ "cudnn_deterministic": false,
46
+ "cudnn_benchmark": false,
47
+ "training_seed": 54321,
48
+ "model": "fast_pitch",
49
+ "num_loader_workers": 0,
50
+ "num_eval_loader_workers": 0,
51
+ "use_noise_augment": false,
52
+ "audio": {
53
+ "fft_size": 1024,
54
+ "win_length": 1024,
55
+ "hop_length": 256,
56
+ "frame_shift_ms": null,
57
+ "frame_length_ms": null,
58
+ "stft_pad_mode": "reflect",
59
+ "sample_rate": 22050,
60
+ "resample": false,
61
+ "preemphasis": 0.0,
62
+ "ref_level_db": 20,
63
+ "do_sound_norm": false,
64
+ "log_func": "np.log",
65
+ "do_trim_silence": true,
66
+ "trim_db": 60.0,
67
+ "do_rms_norm": false,
68
+ "db_level": null,
69
+ "power": 1.5,
70
+ "griffin_lim_iters": 60,
71
+ "num_mels": 80,
72
+ "mel_fmin": 0.0,
73
+ "mel_fmax": 8000,
74
+ "spec_gain": 1.0,
75
+ "do_amp_to_db_linear": true,
76
+ "do_amp_to_db_mel": true,
77
+ "pitch_fmax": 640.0,
78
+ "pitch_fmin": 0.0,
79
+ "signal_norm": false,
80
+ "min_level_db": -100,
81
+ "symmetric_norm": true,
82
+ "max_norm": 4.0,
83
+ "clip_norm": true,
84
+ "stats_path": null
85
+ },
86
+ "use_phonemes": false,
87
+ "phonemizer": null,
88
+ "phoneme_language": "en-us",
89
+ "compute_input_seq_cache": false,
90
+ "text_cleaner": "multilingual_cleaners",
91
+ "enable_eos_bos_chars": false,
92
+ "test_sentences_file": "",
93
+ "phoneme_cache_path": "output_indic_fastpitch/tef13x/phoneme_cache",
94
+ "characters": {
95
+ "characters_class": "TTS.tts.models.vits.VitsCharacters",
96
+ "vocab_dict": null,
97
+ "pad": "<PAD>",
98
+ "eos": "<EOS>",
99
+ "bos": "<BOS>",
100
+ "blank": "<BLNK>",
101
+ "characters": " !',.:;?o\u0c02\u0c03\u0c05\u0c06\u0c07\u0c08\u0c09\u0c0a\u0c0b\u0c0e\u0c0f\u0c10\u0c12\u0c13\u0c14\u0c15\u0c16\u0c17\u0c18\u0c19\u0c1a\u0c1b\u0c1c\u0c1d\u0c1e\u0c1f\u0c20\u0c21\u0c22\u0c23\u0c24\u0c25\u0c26\u0c27\u0c28\u0c2a\u0c2b\u0c2c\u0c2d\u0c2e\u0c2f\u0c30\u0c31\u0c32\u0c33\u0c35\u0c36\u0c37\u0c38\u0c39\u0c3e\u0c3f\u0c40\u0c41\u0c42\u0c43\u0c46\u0c47\u0c48\u0c4a\u0c4b\u0c4c\u0c4d\u201d\ufeff",
102
+ "punctuations": "!\u00a1'(),-.:;\u00bf? ",
103
+ "phonemes": null,
104
+ "is_unique": true,
105
+ "is_sorted": true
106
+ },
107
+ "add_blank": false,
108
+ "batch_group_size": 0,
109
+ "loss_masking": null,
110
+ "sort_by_audio_len": true,
111
+ "min_audio_len": 1,
112
+ "max_audio_len": 441000,
113
+ "min_text_len": 1,
114
+ "max_text_len": 400,
115
+ "compute_f0": true,
116
+ "compute_linear_spec": false,
117
+ "precompute_num_workers": 0,
118
+ "start_by_longest": false,
119
+ "datasets": [
120
+ {
121
+ "name": "indictts",
122
+ "path": "/home/ttsteam/datasets/indictts/tef13x",
123
+ "meta_file_train": "metadata_train.csv",
124
+ "ignored_speakers": null,
125
+ "language": "tef13x",
126
+ "meta_file_val": "metadata_test.csv",
127
+ "meta_file_attn_mask": ""
128
+ }
129
+ ],
130
+ "test_sentences": [
131
+ "\u0c38\u0c3f\u0c02\u0c39\u0c02 \u0c05\u0c21\u0c4d\u0c21\u0c41\u0c35\u0c1a\u0c4d\u0c1a\u0c3f, \u0c24\u0c2a\u0c4d\u0c2a\u0c41\u0c15\u0c4b \u0c36\u0c3f\u0c15\u0c4d\u0c37 \u0c35\u0c3f\u0c27\u0c3f\u0c02\u0c1a\u0c35\u0c32\u0c38\u0c3f\u0c02\u0c26\u0c3f \u0c28\u0c47\u0c28\u0c41 \u0c05\u0c28\u0c3f \u0c15\u0c4b\u0c24\u0c3f\u0c28\u0c3f \u0c05\u0c19\u0c4d\u0c1e\u0c3e\u0c2a\u0c3f\u0c02\u0c1a\u0c3f\u0c02\u0c26\u0c3f \u0c28\u0c15\u0c4d\u0c15\u0c15\u0c47\u0c38\u0c3f \u0c24\u0c3f\u0c30\u0c3f\u0c17\u0c3f \u0c2e\u0c02\u0c24\u0c4d\u0c30\u0c3f \u0c2a\u0c41\u0c02\u0c17\u0c35\u0c3e \u0c08 \u0c2e\u0c42\u0c37\u0c3f\u0c15\u0c3e\u0c27\u0c2e\u0c41\u0c21\u0c41 \u0c1a\u0c4b\u0c30\u0c41\u0c21\u0c41 \u0c05\u0c28\u0c3f \u0c28\u0c40\u0c15\u0c41 \u0c0e\u0c32\u0c3e \u0c24\u0c46\u0c32\u0c3f\u0c38\u0c3f\u0c02\u0c26\u0c3f \u0c05\u0c28\u0c3f \u0c05\u0c21\u0c3f\u0c17\u0c3f\u0c02\u0c26\u0c3f.",
132
+ "\u0c08 \u0c2e\u0c3e\u0c1f\u0c32\u0c41 \u0c35\u0c3f\u0c02\u0c1f\u0c42\u0c28\u0c47 \u0c17\u0c3e\u0c32\u0c35\u0c41\u0c21\u0c41, \u0c15\u0c41\u0c35\u0c32\u0c2f\u0c3e\u0c36\u0c4d\u0c35\u0c3e\u0c28\u0c4d\u0c28\u0c3f \u0c0e\u0c15\u0c4d\u0c15\u0c3f, \u0c36\u0c24\u0c4d\u0c30\u0c41\u0c1c\u0c3f\u0c24\u0c4d\u0c24\u0c41\u0c35\u0c26\u0c4d\u0c26\u0c15\u0c41 \u0c35\u0c46\u0c33\u0c4d\u0c32\u0c3f, \u0c0b\u0c24\u0c41\u0c27\u0c4d\u0c35\u0c1c\u0c41\u0c23\u0c4d\u0c23\u0c3f \u0c2a\u0c02\u0c2a\u0c2e\u0c28\u0c3f \u0c15\u0c4b\u0c30\u0c3e\u0c21\u0c41, \u0c0b\u0c24\u0c41\u0c27\u0c4d\u0c35\u0c1c\u0c41\u0c21\u0c41, \u0c15\u0c41\u0c35\u0c32\u0c2f\u0c3e\u0c36\u0c4d\u0c35\u0c3e\u0c28\u0c4d\u0c28\u0c3f \u0c0e\u0c15\u0c4d\u0c15\u0c3f, \u0c17\u0c3e\u0c32\u0c35\u0c41\u0c21\u0c3f \u0c35\u0c46\u0c02\u0c1f, \u0c06\u0c2f\u0c28 \u0c06\u0c36\u0c4d\u0c30\u0c2e\u0c3e\u0c28\u0c3f\u0c15\u0c3f \u0c35\u0c46\u0c33\u0c4d\u0c33\u0c3e\u0c21\u0c41."
133
+ ],
134
+ "eval_split_max_size": null,
135
+ "eval_split_size": 0.01,
136
+ "use_speaker_weighted_sampler": false,
137
+ "speaker_weighted_sampler_alpha": 1.0,
138
+ "use_language_weighted_sampler": false,
139
+ "language_weighted_sampler_alpha": 1.0,
140
+ "use_length_weighted_sampler": false,
141
+ "length_weighted_sampler_alpha": 1.0,
142
+ "base_model": "forward_tts",
143
+ "model_args": {
144
+ "num_chars": 89,
145
+ "out_channels": 80,
146
+ "hidden_channels": 512,
147
+ "use_aligner": true,
148
+ "use_pitch": true,
149
+ "pitch_predictor_hidden_channels": 256,
150
+ "pitch_predictor_kernel_size": 3,
151
+ "pitch_predictor_dropout_p": 0.1,
152
+ "pitch_embedding_kernel_size": 3,
153
+ "duration_predictor_hidden_channels": 256,
154
+ "duration_predictor_kernel_size": 3,
155
+ "duration_predictor_dropout_p": 0.1,
156
+ "positional_encoding": true,
157
+ "poisitonal_encoding_use_scale": true,
158
+ "length_scale": 1,
159
+ "encoder_type": "fftransformer",
160
+ "encoder_params": {
161
+ "hidden_channels_ffn": 1024,
162
+ "num_heads": 1,
163
+ "num_layers": 6,
164
+ "dropout_p": 0.1
165
+ },
166
+ "decoder_type": "fftransformer",
167
+ "decoder_params": {
168
+ "hidden_channels_ffn": 1024,
169
+ "num_heads": 1,
170
+ "num_layers": 6,
171
+ "dropout_p": 0.1
172
+ },
173
+ "detach_duration_predictor": false,
174
+ "max_duration": 75,
175
+ "num_speakers": 2,
176
+ "use_speaker_embedding": true,
177
+ "speakers_file": "models/v1/te/fastpitch/speakers.pth",
178
+ "use_d_vector_file": false,
179
+ "d_vector_dim": 512,
180
+ "d_vector_file": null,
181
+ "use_speaker_encoder_as_loss": false,
182
+ "speaker_encoder_config_path": "",
183
+ "speaker_encoder_model_path": "",
184
+ "vocoder_path": null,
185
+ "vocoder_config_path": null,
186
+ "use_separate_optimizers": false
187
+ },
188
+ "return_wav": false,
189
+ "num_speakers": 2,
190
+ "speakers_file": "models/v1/te/fastpitch/speakers.pth",
191
+ "use_speaker_embedding": true,
192
+ "use_d_vector_file": false,
193
+ "d_vector_file": "",
194
+ "d_vector_dim": 512,
195
+ "spec_loss_type": "mse",
196
+ "duration_loss_type": "mse",
197
+ "use_ssim_loss": false,
198
+ "ssim_loss_alpha": 1.0,
199
+ "spec_loss_alpha": 1.0,
200
+ "aligner_loss_alpha": 1.0,
201
+ "pitch_loss_alpha": 0.1,
202
+ "dur_loss_alpha": 0.1,
203
+ "binary_align_loss_alpha": 0.1,
204
+ "spk_encoder_loss_alpha": 0.1,
205
+ "binary_loss_warmup_epochs": 150,
206
+ "aligner_epochs": 0,
207
+ "min_seq_len": 13,
208
+ "max_seq_len": 500000,
209
+ "r": 1,
210
+ "f0_cache_path": "output_indic_fastpitch/tef13x/f0_cache"
211
+ }
speakers.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f665e358b34b232fb27f7c8cd3968fcd47784a7be065ae127f611c33ee809bea
3
+ size 431