New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Claim is misleading. CNN is not slower. The CNN implementation is slow. #1

Open

geyang opened this issue Apr 24, 2021 · 1 comment

geyang commented Apr 24, 2021

The claim in this article and the notebook is quite misleading.

pyTorch CNN implementation already uses the cuDNN or a faster fbfft module. For smaller kernel sizes, the Winograd algorithm is even faster.

Your comparison is against convolution done in the position space, using a python for loop.

eduardo4jesus commented Jun 22, 2021

pyTorch CNN implementation already uses the cuDNN or a faster fbfft module

@geyang. Could you link that code from pytorch's github? I am very interested in that and I couldn't find that yet.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment