benchmark

0ffd31ab - use larger multiplier for smaller tensors for a few models (#132952)

use larger multiplier for smaller tensors for a few models (#132952) Summary: Fix https://github.com/pytorch/pytorch/issues/132922 and https://github.com/pytorch/pytorch/issues/132924 X-link: https://github.com/pytorch/pytorch/pull/132952 Approved by: https://github.com/eellison, https://github.com/jansel Reviewed By: PaliC Differential Revision: D61000058 Pulled By: shunting314 fbshipit-source-id: cd4daa111f882336b03f49f91beec6897f0130d4