Hi, since pytorch installation does let you choose between cuda or ROCm usage only and not both, I’m writing this post. How can I use in the same code an amd and nvidia gpu simultaneously in the same computer ? Is this feature already implemented ? I want to experiment and play with distributed inference.