pytorch-bayes-by-backprop icon indicating copy to clipboard operation
pytorch-bayes-by-backprop copied to clipboard

Published 20 hours ago • verified publisher icon mjpyeon

Calling forward after the backward pass and parameter update

Open rohanpaleja27 opened this issue 5 years ago • 0 comments

Hi,

Thank you for keeping your code open source. While going through it, I see that you are calling a forward pass after your parameter update, which makes the printed training accuracy incorrect.

If there is a reason for this, please let me know as I am not highly familiar with BNNs.

Best,

rohanpaleja27 avatar May 16 '19 19:05 rohanpaleja27