CUDA_KERNEL_LOOP: prevent int overflow in loop increment. (#24818)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/24309.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/24818
Differential Revision: D16927215
Pulled By: ezyang
fbshipit-source-id: aeab5226fec6045941399693479975db4542c79e