Switch empty_like to use merge_in to process TensorOptions. (#33505)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/33505
This shouldn't change semantics, but it has the benefit of making
torch::empty_like(x, dtype(kFloat)) actually work (previously, this
would just ignore all of the properties from x).
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Differential Revision: D20001776
Pulled By: ezyang
fbshipit-source-id: ba81186d3293abc65da6130b2684d42e9e675208