[TensorExpr] Distinguish aten::max reduction op from aten::max elementwise op and only fuse the latter. (#38171)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/38171
Test Plan: Imported from OSS
Differential Revision: D21487389
Pulled By: ZolotukhinM
fbshipit-source-id: ac28789bf2bea389f560de4d5b979e036295e96a
Author
Mikhail Zolotukhin