Update from ori#1

Merged

tarsbase merged 138 commits intotarsbase:masterfrom

fatchord:master

Sep 2, 2019

Commits on Mar 24, 2019

Commits on Mar 25, 2019

Add pretrained model link
fatchord
authored

Commits on Mar 26, 2019

Commits on Mar 27, 2019

Prevent overrun in preprocessing
fatchord
committed

Commits on Mar 28, 2019

Commits on Apr 2, 2019

Commits on Apr 3, 2019

Add dependencies
fatchord
committed

Commits on Apr 5, 2019

Commits on Apr 7, 2019

Commits on Apr 8, 2019

Commits on Apr 30, 2019

Update Readme
fatchord
committed

Commits on May 3, 2019

Add training animation
fatchord
committed

Commits on May 6, 2019

Commits on Jul 11, 2019

Commits on Jul 14, 2019

Merge pull request #107 from TheButlah/master
fatchord
authored

Commits on Jul 18, 2019

Enabled multi-gpu training, buffers, grad clip in vocoder
TheButlah
committed
Made models `num_params()` actually useful if `print_out==False`
TheButlah
committed
Fixed a bunch of my own bugs, now saving optimizer state

TheButlah
authored and
Ryan Butler
committed

Commits on Jul 19, 2019

Fixed bug due to model.load() placing tensors on CPU.

TheButlah
authored and
Ryan Butler
committed
Fixed bug with improper comparison `device == torch.device('cuda')`

TheButlah
authored and
Ryan Butler
committed

Commits on Jul 20, 2019

Fixed warnings for flattening parameters in rnns
TheButlah
committed

Commits on Jul 23, 2019

Fixed missing import, added type annotation for IDE suggestions

TheButlah
authored and
Ryan Butler
committed
Fixed bugs in multi-gpu tacotron where tensors were on wrong device

TheButlah
authored and
Ryan Butler
committed

Commits on Jul 24, 2019

Made preprocess.py use a configurable number of workers

TheButlah
authored and
Ryan Butler
committed
Made self.r in Tacotron default to 1

TheButlah
authored and
Ryan Butler
committed
Merge pull request #109 from TheButlah/master
fatchord
authored

Commits on Jul 25, 2019