Closed Yeongtae closed 5 years ago
Thanks. I've updated to the latest version of Tacotron-2 and forked this version. Currently training the LJ taco model.
Thanks you both. I am looking forward to your progress.
WaveRNN use GRU operation optimization techniques. To figure out efficiency of operation optimization, I had completed the GRU opreation optimization experiment.
I had compared Pythorch's built-in GRU module with self implementation of GRU operation optimization
I made a simple two-dimensional coordinate prediction problem, f(xi) = y = x(i+1), The dataset was generated by using r x rotationmatrix(theta).[1, 0 ]^T. Device: k80 GPU
Built-in GRU module Self implementation of original GRU Self implementation of GRU operation optimization
I performed full batch inferences, 10000 times each.
Built-in GRU module: 6.02 seconds Self implementation of original GRU: 16.0 seconds Self implementation of GRU operation optimization: 9.70 seconds
The pytorch build-in GRU module is well optimized. Because it use cudnn_gru, check https://github.com/rossumai/OCkRE/issues/2
In my opinion, it's faster to use built-in GRU in pytorch.
Thanks for the experiment. Unfortunately I'm in hospital and couldn't continue to integrate the original wavernn implementation for now, but it's still on my agenda.
I'm trying to use faster wave generator, which is waveglow. Sorry about stopping implementation.
Perhaps this post is interesting for you: https://github.com/mozilla/TTS/issues/9#issuecomment-440976514 I'm generally still watching the Mozilla TTS repo to check their progress with WaveRNN.
I'll close this for now.
Data feeding
[ ] using parameters in hparam.py to reduce independent parameters (3rd)[ ] muLaw quantization to preserve more important information (4th)Model
modifying the wavernn model to reduce matrix multiflications (1st, ~ 11.04) (Fail)11.1111.14)[ ] weight sparsification to reduce computation times (2nd, ~11.16)