It didn't perform any better than the regular model Removing batch normalisation significantly harmed training performance
384 B
384 B
It didn't perform any better than the regular model Removing batch normalisation significantly harmed training performance