pytorch

e838137b - Add high level control of fp32 matmul precision; disable TF32 for matmuls by default

Add high level control of fp32 matmul precision; disable TF32 for matmuls by default #76440 CC @mruberry @ptrblck Pull Request resolved: https://github.com/pytorch/pytorch/pull/76509 Approved by: https://github.com/ngimel