Fix inaccurate dispatch tables (#54127)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/54127
During the meta tensor bringup, I found all of these operators
advertised that they worked on all backends (DefaultBackend/Math)
but actually they only worked on CPU/CUDA.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Reviewed By: mruberry
Differential Revision: D27109508
Pulled By: ezyang
fbshipit-source-id: 0f474ecf4aba8b8207f2910bdc962bf581f53853