Hi all,
When I run the model that is based on nn.linear, the server shows ~2000% cpu usage, I do not mind that, however, when I used nn.Conv2D I noticed the cpu is 100% maximum, and the iterations/epochs are extremely slow.
I am wondering why is that, is there is a default parallel option for nn.linear, and is it possible to enable such thing for nn.Conv2D()?
I appreciate your help.