[RFC]fix test_ddp_logging_data_cpu with tsan (#54465)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/54465
It is reported that there is data race issue when the test runs with tsan. The root cause is from 'model.frc1.double()' call. This is not because DistributedDataParallel() works together with 'model.frc1.double()'. If we remove DistributedDataParallel(), just call 'model.frc1.double(); model.frc2.double();', it complained the same data race issue.
I'm not sure how to do data type cast in this test without tsan complains, so removing this line of codes and mixed data type logging check.
Please kindly let me know if you have a better suggestion on how to do data type cast correctly
Test Plan: unit test
Reviewed By: SciPioneer
Differential Revision: D27249821
fbshipit-source-id: 0368157e11cbe7d15828dccca78271d89d502ec4
