ResGAN icon indicating copy to clipboard operation
ResGAN copied to clipboard

Published 20 hours ago • verified publisher icon ljuvela

TensorFlow implementation with WGAN

Open ljuvela opened this issue 6 years ago • 3 comments

Everything is so much nicer in TensorFlow. No CPU/GPU memory swapping for FFTs, for example.

Stay tuned.

ljuvela avatar May 14 '18 09:05 ljuvela

when I run the projec,it shows errors:('cannt convert tensor to tensortype') model.py line 37 in theano_fft frame=T.set_subtensor(frame[:,:400],x_win) i confused so much. please help me to figure it out. thank you so so much.

qinshan2017 avatar Jul 26 '18 03:07 qinshan2017

That's probably some issue with old Theano versions, which I'd rather get rid of entirely. I'm now testing a Keras + Tensorflow version for the code, and that will probably remove your problem.

I'll try to push the fix later today. Thanks for the reminder.

ljuvela avatar Jul 26 '18 08:07 ljuvela

The Tensorflow FFT is now in the repo and all Theano requirements have been removed.

ljuvela avatar Jul 27 '18 09:07 ljuvela