Skip to content

Disabling gradient tracking (torch.no_grad())

Jump to bottom
Corey Lowman edited this page Aug 14, 2022 · 2 revisions

All you have to do is not insert a tape into the input tensor. I.e. DON'T call .trace()

See Is-NoTape-&-NoneTape-equivalent-to-torch.no_grad() for more info.

Clone this wiki locally