News for hackers. Stuff that matters.
Re:Fast, GPU-native differentiable physics and beyond with T... by dionhaefner on Friday October 30, 2026 @09:25AM (#50303561)
We thought it would be cool to be able to run GPU workloads across process boundaries from JAX and torch. Train your torch networks through Julia, C++ solvers with adjoints, or any arcane autodiff potion you choose to concoct – now certified fresh without host round trips. Bon apetit!