delete brodcasting ops from shape analysis resize aliasing (#20661)
Summary:
According to https://pytorch.org/docs/stable/notes/broadcasting.html, in-place operations do not allow the in-place tensor to change shape as a result of the broadcast. Therefore our shape analysis could keep the shape information on inputs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20661
Differential Revision: D15406477
Pulled By: wanchaol
fbshipit-source-id: 8ab60e783292f2fe26e5fdecfb64bec43bca6826