pytorch

c78a4a52 - remove unnecessary/dead code in upsample_nearest1d cuda kernel (#51916)

remove unnecessary/dead code in upsample_nearest1d cuda kernel (#51916) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/51916 After getting ported to structured kernels, the vector overloads of `upsample_nearest1d` are DefaultBackend kernels, meaning they are backend agnostic. We can kill their CUDA-specific implementations. I also removed a few redundant checks in the cuda kernels that are now performed by the meta shape-checking function. Test Plan: Imported from OSS Reviewed By: ezyang Differential Revision: D26327749 Pulled By: bdhirsh fbshipit-source-id: b5a17e14237fb36236d4079433f99c71cd3beef3