Fix TestOpInfoCUDA.test_unsupported_dtypes_addmm_cuda_bfloat16 on ampere (#50440)
Summary:
The `TestOpInfoCUDA.test_unsupported_dtypes_addmm_cuda_bfloat16` in `test_ops.py` is failing on ampere. This is because addmm is supported on Ampere, but the test is asserting that it is not supported.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/50440
Reviewed By: mrshenli
Differential Revision: D25893326
Pulled By: ngimel
fbshipit-source-id: afeec25fdd76e7336d84eb53ea36319ade1ab421
