Siddhant commited on
Commit
de4afde
1 Parent(s): cdcc9f5

import from zenodo

Browse files
Files changed (18) hide show
  1. README.md +50 -0
  2. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/config.yaml +253 -0
  3. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/backward_time.png +0 -0
  4. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/duration_loss.png +0 -0
  5. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/energy_loss.png +0 -0
  6. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/forward_time.png +0 -0
  7. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/iter_time.png +0 -0
  8. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/l1_loss.png +0 -0
  9. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/loss.png +0 -0
  10. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/lr_0.png +0 -0
  11. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/optim_step_time.png +0 -0
  12. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/pitch_loss.png +0 -0
  13. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/train_time.png +0 -0
  14. exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/train.loss.ave_5best.pth +3 -0
  15. exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/energy_stats.npz +0 -0
  16. exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/feats_stats.npz +0 -0
  17. exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/pitch_stats.npz +0 -0
  18. meta.yaml +8 -0
README.md ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - espnet
4
+ - audio
5
+ - text-to-speech
6
+ language: ja
7
+ datasets:
8
+ - jsut
9
+ license: cc-by-4.0
10
+ ---
11
+ ## Example ESPnet2 TTS model
12
+ ### `kan-bayashi/jsut_conformer_fastspeech2`
13
+ ♻️ Imported from https://zenodo.org/record/4032246/
14
+
15
+ This model was trained by kan-bayashi using jsut/tts1 recipe in [espnet](https://github.com/espnet/espnet/).
16
+ ### Demo: How to use in ESPnet2
17
+ ```python
18
+ # coming soon
19
+ ```
20
+ ### Citing ESPnet
21
+ ```BibTex
22
+ @inproceedings{watanabe2018espnet,
23
+ author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson {Enrique Yalta Soplin} and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
24
+ title={{ESPnet}: End-to-End Speech Processing Toolkit},
25
+ year={2018},
26
+ booktitle={Proceedings of Interspeech},
27
+ pages={2207--2211},
28
+ doi={10.21437/Interspeech.2018-1456},
29
+ url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
30
+ }
31
+ @inproceedings{hayashi2020espnet,
32
+ title={{Espnet-TTS}: Unified, reproducible, and integratable open source end-to-end text-to-speech toolkit},
33
+ author={Hayashi, Tomoki and Yamamoto, Ryuichi and Inoue, Katsuki and Yoshimura, Takenori and Watanabe, Shinji and Toda, Tomoki and Takeda, Kazuya and Zhang, Yu and Tan, Xu},
34
+ booktitle={Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
35
+ pages={7654--7658},
36
+ year={2020},
37
+ organization={IEEE}
38
+ }
39
+ ```
40
+ or arXiv:
41
+ ```bibtex
42
+ @misc{watanabe2018espnet,
43
+ title={ESPnet: End-to-End Speech Processing Toolkit},
44
+ author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Enrique Yalta Soplin and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
45
+ year={2018},
46
+ eprint={1804.00015},
47
+ archivePrefix={arXiv},
48
+ primaryClass={cs.CL}
49
+ }
50
+ ```
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/config.yaml ADDED
@@ -0,0 +1,253 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ config: conf/tuning/train_conformer_fastspeech2.yaml
2
+ print_config: false
3
+ log_level: INFO
4
+ dry_run: false
5
+ iterator_type: sequence
6
+ output_dir: exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk
7
+ ngpu: 1
8
+ seed: 0
9
+ num_workers: 1
10
+ num_att_plot: 3
11
+ dist_backend: nccl
12
+ dist_init_method: env://
13
+ dist_world_size: null
14
+ dist_rank: null
15
+ local_rank: 0
16
+ dist_master_addr: null
17
+ dist_master_port: null
18
+ dist_launcher: null
19
+ multiprocessing_distributed: false
20
+ cudnn_enabled: true
21
+ cudnn_benchmark: false
22
+ cudnn_deterministic: true
23
+ collect_stats: false
24
+ write_collected_feats: false
25
+ max_epoch: 1000
26
+ patience: null
27
+ val_scheduler_criterion:
28
+ - valid
29
+ - loss
30
+ early_stopping_criterion:
31
+ - valid
32
+ - loss
33
+ - min
34
+ best_model_criterion:
35
+ - - valid
36
+ - loss
37
+ - min
38
+ - - train
39
+ - loss
40
+ - min
41
+ keep_nbest_models: 5
42
+ grad_clip: 1.0
43
+ grad_clip_type: 2.0
44
+ grad_noise: false
45
+ accum_grad: 10
46
+ no_forward_run: false
47
+ resume: true
48
+ train_dtype: float32
49
+ use_amp: false
50
+ log_interval: null
51
+ pretrain_path: []
52
+ pretrain_key: []
53
+ num_iters_per_epoch: 500
54
+ batch_size: 20
55
+ valid_batch_size: null
56
+ batch_bins: 2400000
57
+ valid_batch_bins: null
58
+ train_shape_file:
59
+ - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/text_shape.phn
60
+ - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/speech_shape
61
+ valid_shape_file:
62
+ - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/valid/text_shape.phn
63
+ - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/valid/speech_shape
64
+ batch_type: numel
65
+ valid_batch_type: null
66
+ fold_length:
67
+ - 150
68
+ - 240000
69
+ sort_in_batch: descending
70
+ sort_batch: descending
71
+ multiple_iterator: false
72
+ chunk_length: 500
73
+ chunk_shift_ratio: 0.5
74
+ num_cache_chunks: 1024
75
+ train_data_path_and_name_and_type:
76
+ - - dump/raw/tr_no_dev/text
77
+ - text
78
+ - text
79
+ - - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/tr_no_dev/durations
80
+ - durations
81
+ - text_int
82
+ - - dump/raw/tr_no_dev/wav.scp
83
+ - speech
84
+ - sound
85
+ - - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/collect_feats/pitch.scp
86
+ - pitch
87
+ - npy
88
+ - - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/collect_feats/energy.scp
89
+ - energy
90
+ - npy
91
+ valid_data_path_and_name_and_type:
92
+ - - dump/raw/dev/text
93
+ - text
94
+ - text
95
+ - - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/dev/durations
96
+ - durations
97
+ - text_int
98
+ - - dump/raw/dev/wav.scp
99
+ - speech
100
+ - sound
101
+ - - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/valid/collect_feats/pitch.scp
102
+ - pitch
103
+ - npy
104
+ - - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/valid/collect_feats/energy.scp
105
+ - energy
106
+ - npy
107
+ allow_variable_data_keys: false
108
+ max_cache_size: 0.0
109
+ valid_max_cache_size: null
110
+ optim: adam
111
+ optim_conf:
112
+ lr: 1.0
113
+ scheduler: noamlr
114
+ scheduler_conf:
115
+ model_size: 384
116
+ warmup_steps: 4000
117
+ token_list:
118
+ - <blank>
119
+ - <unk>
120
+ - ty
121
+ - dy
122
+ - v
123
+ - py
124
+ - my
125
+ - by
126
+ - ny
127
+ - hy
128
+ - gy
129
+ - ry
130
+ - ky
131
+ - f
132
+ - p
133
+ - z
134
+ - ch
135
+ - ts
136
+ - j
137
+ - b
138
+ - y
139
+ - h
140
+ - cl
141
+ - I
142
+ - U
143
+ - w
144
+ - g
145
+ - d
146
+ - sh
147
+ - pau
148
+ - m
149
+ - N
150
+ - s
151
+ - r
152
+ - t
153
+ - n
154
+ - k
155
+ - e
156
+ - u
157
+ - i
158
+ - o
159
+ - a
160
+ - <sos/eos>
161
+ odim: null
162
+ model_conf: {}
163
+ use_preprocessor: true
164
+ token_type: phn
165
+ bpemodel: null
166
+ non_linguistic_symbols: null
167
+ cleaner: jaconv
168
+ g2p: pyopenjtalk
169
+ feats_extract: fbank
170
+ feats_extract_conf:
171
+ fs: 24000
172
+ fmin: 80
173
+ fmax: 7600
174
+ n_mels: 80
175
+ hop_length: 300
176
+ n_fft: 2048
177
+ win_length: 1200
178
+ normalize: global_mvn
179
+ normalize_conf:
180
+ stats_file: exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/feats_stats.npz
181
+ tts: fastspeech2
182
+ tts_conf:
183
+ adim: 384
184
+ aheads: 2
185
+ elayers: 4
186
+ eunits: 1536
187
+ dlayers: 4
188
+ dunits: 1536
189
+ positionwise_layer_type: conv1d
190
+ positionwise_conv_kernel_size: 3
191
+ duration_predictor_layers: 2
192
+ duration_predictor_chans: 256
193
+ duration_predictor_kernel_size: 3
194
+ postnet_layers: 5
195
+ postnet_filts: 5
196
+ postnet_chans: 256
197
+ use_masking: true
198
+ encoder_normalize_before: false
199
+ decoder_normalize_before: false
200
+ reduction_factor: 1
201
+ encoder_type: conformer
202
+ decoder_type: conformer
203
+ conformer_pos_enc_layer_type: rel_pos
204
+ conformer_self_attn_layer_type: rel_selfattn
205
+ conformer_activation_type: swish
206
+ use_macaron_style_in_conformer: true
207
+ use_cnn_in_conformer: true
208
+ conformer_enc_kernel_size: 7
209
+ conformer_dec_kernel_size: 31
210
+ init_type: xavier_uniform
211
+ transformer_enc_dropout_rate: 0.2
212
+ transformer_enc_positional_dropout_rate: 0.2
213
+ transformer_enc_attn_dropout_rate: 0.2
214
+ transformer_dec_dropout_rate: 0.2
215
+ transformer_dec_positional_dropout_rate: 0.2
216
+ transformer_dec_attn_dropout_rate: 0.2
217
+ pitch_predictor_layers: 5
218
+ pitch_predictor_chans: 256
219
+ pitch_predictor_kernel_size: 5
220
+ pitch_predictor_dropout: 0.5
221
+ pitch_embed_kernel_size: 1
222
+ pitch_embed_dropout: 0.0
223
+ stop_gradient_from_pitch_predictor: true
224
+ energy_predictor_layers: 2
225
+ energy_predictor_chans: 256
226
+ energy_predictor_kernel_size: 3
227
+ energy_predictor_dropout: 0.5
228
+ energy_embed_kernel_size: 1
229
+ energy_embed_dropout: 0.0
230
+ stop_gradient_from_energy_predictor: false
231
+ pitch_extract: dio
232
+ pitch_extract_conf:
233
+ fs: 24000
234
+ n_fft: 2048
235
+ hop_length: 300
236
+ f0max: 400
237
+ f0min: 80
238
+ pitch_normalize: global_mvn
239
+ pitch_normalize_conf:
240
+ stats_file: exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/pitch_stats.npz
241
+ energy_extract: energy
242
+ energy_extract_conf:
243
+ fs: 24000
244
+ n_fft: 2048
245
+ hop_length: 300
246
+ win_length: 1200
247
+ energy_normalize: global_mvn
248
+ energy_normalize_conf:
249
+ stats_file: exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/energy_stats.npz
250
+ required:
251
+ - output_dir
252
+ - token_list
253
+ distributed: false
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/backward_time.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/duration_loss.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/energy_loss.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/forward_time.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/iter_time.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/l1_loss.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/loss.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/lr_0.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/optim_step_time.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/pitch_loss.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/images/train_time.png ADDED
exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/train.loss.ave_5best.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3dff172d7b3ba76d5534731c44b06c09eb82cf9e37052914373573d6a097f2a
3
+ size 281515829
exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/energy_stats.npz ADDED
Binary file (770 Bytes). View file
exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/feats_stats.npz ADDED
Binary file (1.4 kB). View file
exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/stats/train/pitch_stats.npz ADDED
Binary file (770 Bytes). View file
meta.yaml ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
1
+ espnet: 0.8.0
2
+ files:
3
+ model_file: exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/train.loss.ave_5best.pth
4
+ python: "3.7.3 (default, Mar 27 2019, 22:11:17) \n[GCC 7.3.0]"
5
+ timestamp: 1600239314.910561
6
+ torch: 1.6.0
7
+ yaml_files:
8
+ train_config: exp/tts_train_conformer_fastspeech2_raw_phn_jaconv_pyopenjtalk/config.yaml