Stop using dispatchTypeId to do checks for tensor list unwrap. (#32787)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/32787
Gets rid of a longstanding TODO. TensorList unwrap is only used for cat, which
means we can assume that the inputs are dense, and do something similar to how
we do the dense tensor wrapping above.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Differential Revision: D19628642
Pulled By: ezyang
fbshipit-source-id: 3264439407585fb97995a9a2302c2913efecb421