pytorch

51ea4418 - Upcast to fp32 in test_addmm_block ref_half_bfloat16 (#86682)

Upcast to fp32 in test_addmm_block ref_half_bfloat16 (#86682) Fixes https://github.com/pytorch/pytorch/issues/86681 Pull Request resolved: https://github.com/pytorch/pytorch/pull/86682 Approved by: https://github.com/nikitaved