Hi team, is there way to reload when "RuntimeError: CUDA error: device-side assert triggered" ? I know it happens when there are channel size mismatch but want to prevent just one wrong input to mess up whole docker

Maybe you can put an exception block to catch this exception