mozilla
/
TTS
зеркало из https://github.com/mozilla/TTS.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность
643 коммитов 13 Ветки 1 Тег 128 MiB
Граф коммитов

208 Коммитов

Автор SHA1 Сообщение Дата
Eren Golge e2d9b0e27c Merge branch 'dev' 2019-02-18 13:08:47 +01:00
Eren Golge 4d2c374cfd Filter out shape mismatch laeyrs in restoriong model for finetunning 2019-02-16 03:19:26 +01:00
Eren Golge 6ea31e47df Constant queue size for autoregression window 2019-02-16 03:18:49 +01:00
Eren Golge 4fad1971cd set learning rate for restored model 2019-02-12 10:04:39 +01:00
Eren Golge d03865ed05 configurable loss weigting for lower frequencies 2019-02-12 10:04:18 +01:00
Eren Golge 0308c92c16 pick best model by loss function is run_eval is true 2019-02-05 11:56:44 +01:00
Eren Golge b011dafbab pass num_chars in train.py 2019-01-21 14:52:40 +01:00
Eren Golge 7e020d4084 Bug fixes 2019-01-16 16:23:04 +01:00
Eren Golge 72341ae22a phoneme usage bug fix 2019-01-16 13:09:47 +01:00
Eren Golge 5733eab90b Add configurable phoneme cache path 2019-01-16 12:33:24 +01:00
Eren Golge 656542db93 Debug purpose alignment image saving 2019-01-16 12:18:54 +01:00
Eren Golge c7ee056f21 Merge branch 'master' of github.com:mozilla/TTS 2019-01-06 18:13:54 +01:00
Yves-Noel Weweler 10524b885c
Fix NameError: name 'model_dict' is not defined 
Closes #91
 2018-12-31 13:29:39 +01:00
Eren Golge f5c972dee6 bug fixes 2018-12-28 14:22:57 +01:00
Eren Golge 17b65d5cde bug fixes on train.py 2018-12-28 14:22:40 +01:00
Eren Golge 6d8b2dad91 bug fix 2018-12-28 14:18:33 +01:00
Eren Golge a1fe867371 bug fix for partial model initialization, if model is not initialized, it is tried to init model partially with only matching layers in size 2018-12-28 14:18:31 +01:00
Eren Golge 481105ccfa logger for tensorboard plotting 2018-12-28 14:18:19 +01:00
Bajibabu Bollepalli dce5a05e80
Update train.py 2018-12-18 15:49:24 +02:00
Eren Golge 8d865629a0 partial model initialization 2018-12-12 15:43:58 +01:00
Eren Golge 619c73f0f1 Init data_loaders by function beginning of each train and eval run 2018-12-11 17:52:43 +01:00
Eren Golge 1ac037d780 Update README and comments on argsparse 2018-12-11 15:50:58 +01:00
Eren Golge 2ec55e2369 print avg spec length in training 2018-12-11 15:08:02 +01:00
Eren Golge bb2a88a984 Rename LR scheduler 2018-11-26 14:09:42 +01:00
Eren Golge c98631fe36 bug fix 2018-11-24 17:23:46 +01:00
Eren Golge 36efaa789d bug fix in testing phase 2018-11-21 15:42:10 +01:00
Eren Golge 660f8c7b66 print average text length, fix for Nancy preprocessor 2018-11-20 12:54:33 +01:00
Eren Golge 6550db5251 Formatting, fixing import statements, logging learning rate, remove optimizer restore cuda call 2018-11-05 14:05:04 +01:00
Eren Golge bb04a1c6e0 optional lr schedule 2018-11-03 19:47:28 +01:00
Eren Golge c8a552e627 Batch update after data-loss 2018-11-02 16:13:51 +01:00
Eren 1ae3b3e442 Enable CPU training and fix restore_epoch 2018-10-25 14:05:27 +02:00
Eren e00c3f5044 Fix the restore_epoch 2018-10-25 13:56:33 +02:00
Eren 4427067c0b restore lr scheduler 2018-09-29 01:46:11 +02:00
Eren 231607a546 Reduce batch group size 2018-09-26 16:52:47 +02:00
Eren c89a3098dd Merge branch 'weight-decay' into attn-smoothing-bgs-sigmoid-wd 2018-09-26 16:49:39 +02:00
Eren c36a1f7deb Make audio folder and save audio with scipy 2018-09-21 17:41:11 +02:00
Eren 34eeaee58b Make audio folder and save audio with scipy 2018-09-21 17:38:55 +02:00
Eren 30fea0b957 Batch group shuffling 2018-09-20 11:31:39 +02:00
Eren 16db5159f1 Weight decay described here: http://www.fast.ai/2018/07/02/adam-weight-decay/ 2018-09-19 15:46:25 +02:00
Eren 9b29b4e281 Weight decay 2018-09-19 15:46:25 +02:00
Eren 4b08c3a13f Save audio samples per validation 2018-09-16 13:32:59 +02:00
Eren 56c6d0cac8 Remove min max mel freq 2018-09-06 15:26:20 +02:00
Eren 7d66bdc5f4 update model size to paper 2018-09-06 14:37:19 +02:00
Eren 03e1c06f5b bug fix 2018-08-13 15:04:00 +02:00
Eren 73c3ab546d Testing update 2018-08-13 15:02:30 +02:00
Eren bb526c296f Change scheduler AnnealLR and catch audio synthesis error in eval time 2018-08-13 13:13:45 +02:00
Eren 6818e11185 Make lr scheduler configurable 2018-08-12 15:02:06 +02:00
Eren f7add3c8e5 tensorboardx plotting figures 2018-08-11 16:53:09 +02:00
Eren b65f5bd1d0 bug fix for plotting 2018-08-11 16:09:47 +02:00
Eren 96e2e3c776 Use step wise LR scheduler + adapt train.py for passing squence mask directly 2018-08-10 17:48:19 +02:00