pytorch

2cd27be5 - Fix CUDA device guard usage when first arg of kernel is scalar (#39870)

Fix CUDA device guard usage when first arg of kernel is scalar (#39870) Summary: Add an OptionalDeviceGuard for second arg in gpu_kernel_with_scalars when first arg is scalar Closes https://github.com/pytorch/pytorch/issues/38889 Pull Request resolved: https://github.com/pytorch/pytorch/pull/39870 Differential Revision: D22011184 Pulled By: ngimel fbshipit-source-id: 427291c456e879f25d15ab76a60b5d4ad61f3b3f