insert_prepack_unpack pattern (#27102)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/27102
We need to prepack the quantized weight rather then original weight
Test Plan:
.
Imported from OSS
Differential Revision: D17678264
fbshipit-source-id: 50614b841cc41007affcf3df7251f042a5a97c10
