News for hackers. Stuff that matters.

A comment on Copying data is wasteful, mutating data is dangerous. Parent comment

Re:Copying data is wasteful, mutating data is dangerous by markus92 on Friday January 10, 2020 @07:40AM (#22010524)
Do you have any links to some examples or help on how to do this exactly?

Reason being that in our group we have a ton of normalization code that looks exactly like the 'bad' example from this post. As we use relatively large datasets (4D MRI images), memory is a bit scarce...

[ Parent | on HN ]
Re:Copying data is wasteful, mutating data is dangerous by naveen99 on Friday January 10, 2020 @08:03AM (#22010630)
Pytorch has a built in normalize function: https://pytorch.org/docs/master/_modules/torchvision/transfo... It just subtracts the mean and divides by standard deviation, both in place.
[ Parent | on HN ]