Commit Graph

1679 Commits

Author SHA1 Message Date
erogol 55230bb317 plot spectrogram transpose fix 2020-06-16 18:53:48 +02:00
erogol b2e9c05acc losses bug fix 2020-06-16 13:26:16 +02:00
erogol 6948e4ce38 change sample data 2020-06-16 12:35:28 +02:00
erogol 0de38c2617 fixing naming convention in vocoder losses 2020-06-16 12:34:10 +02:00
erogol f18f6e6d3e plot spectrograms undenormalized if no ap is provided 2020-06-16 12:33:36 +02:00
erogol f3d4d1f831 bug fix init AudioProcessor in train.py 2020-06-15 19:26:26 +02:00
erogol e0c44e2cfa Merge branch 'dev' of https://github.com/mozilla/TTS into dev 2020-06-15 19:10:45 +02:00
erogol c90145eef9 add feature path to vocder config 2020-06-15 19:04:05 +02:00
erogol 460d0675a6 bug fix for vocoder feature macthing loss 2020-06-15 19:04:05 +02:00
erogol 98fd3b53f2 load mel features if 'feature_path' is provided 2020-06-15 19:04:05 +02:00
erogol 3c20afa1c9 do not call .item() if returned loss is not torch type 2020-06-15 19:04:05 +02:00
erogol 8f72ad900a remove redundant params in vocoder config file 2020-06-15 19:04:05 +02:00
erogol 5547ddeb61 update vocoder loss implemenatations and fix MSEDLoss 2020-06-15 19:04:05 +02:00
erogol 369a912914 more comments to audio.py 2020-06-15 19:04:05 +02:00
erogol bc69a54f44 visualization bug fix 2020-06-15 19:04:05 +02:00
erogol b62ad06f4c bug fix 2020-06-15 19:04:05 +02:00
erogol 28ee6af06c enable loading precomputed vocoder dataset 2020-06-15 19:04:05 +02:00
erogol 9e71d4d66d reformatting vocoder training and disable G_adv eval if step < d_training 2020-06-15 19:04:05 +02:00
erogol b8766c05ce bug fix 2020-06-15 19:04:05 +02:00
erogol e8a59db839 linter fix 2020-06-15 19:04:05 +02:00
erogol ddfede1f52 restore vocoder scheduler if it is in checkpoint 2020-06-15 19:04:05 +02:00
erogol b3efbbf54b bug fix 2020-06-15 19:04:05 +02:00
erogol 0a8e4cf9d6 new config argument to plot tb figures at training 2020-06-15 19:04:05 +02:00
erogol 9b15331eeb better stats logging for TTS training 2020-06-15 19:04:05 +02:00
erogol 8a931e705c q 2020-06-15 19:04:05 +02:00
erogol 8a201c92c6 multiband melga nconfig 2020-06-15 19:04:05 +02:00
erogol 0fa17bd9bb save schedulers with checkpoints 2020-06-15 19:04:05 +02:00
erogol ecff45369b print lrg and lrD 2020-06-15 19:04:05 +02:00
erogol 74e6e29479 add lr schedulers for generator and discriminator 2020-06-15 19:04:05 +02:00
erogol 1dfafe003d fix multiband melgan inference 2020-06-15 19:04:05 +02:00
erogol c866f23af6 remove weight norm for melgan generator 2020-06-15 19:04:05 +02:00
erogol f9698af06e correction for vocoder configs 2020-06-15 19:04:05 +02:00
erogol 0245a2e258 change alignment_score to alignment_error 2020-06-15 19:04:05 +02:00
erogol 9913724a70 TTS config update for using DDC 2020-06-15 19:04:05 +02:00
erogol f0616a3b24 linter fix 2020-06-15 19:04:05 +02:00
erogol 55b8952c3d run D in eval 2020-06-15 19:04:05 +02:00
erogol 78d43664a1 bug fix and var renaming 2020-06-15 19:04:05 +02:00
erogol 1daba4371d bug fix vocoder model restore 2020-06-15 19:04:05 +02:00
erogol b736c6cdac bug fix for TTS model restore 2020-06-15 19:04:05 +02:00
erogol d2cfab71d3 bug fix for DDC model eval run 2020-06-15 19:04:05 +02:00
erogol 2b377b5487 model config update 2020-06-15 19:04:05 +02:00
erogol d7022a33b3 correct name 2020-06-15 19:04:05 +02:00
erogol 06d3eb42cb bug fix for rwd discriminator 2020-06-15 19:04:05 +02:00
erogol acb367be26 bug fix vocoder training 2020-06-15 19:04:05 +02:00
erogol 16eb15a5ff lint updates 2020-06-15 19:04:05 +02:00
erogol 9f5f3910e4 check DDC parameters in config.json 2020-06-15 19:04:05 +02:00
erogol 01339e94a5 setup_model update following TacotronAbstract 2020-06-15 19:04:05 +02:00
erogol 4aaf57b50b train.py update for DDC 2020-06-15 19:04:05 +02:00
erogol fbf1689d18 update model tests for ddc 2020-06-15 19:04:05 +02:00
erogol 5962515581 update dummy model config 2020-06-15 19:04:05 +02:00