pytorch/torch/distributed/algorithms
Yi Wang b4cb020c0f [Gradient Compression] Make orthogonalization_epsilon configurable in PowerSGDState (#55738)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/55738

Per title, and use 0 as the default value.

It turns out that setting this epsilon as 0 can accelerate convergence and improve accuracy for some use cases.

Test Plan:
unit tests
f264687105
f264675194

Reviewed By: shuyingsunshine21

Differential Revision: D27694971

fbshipit-source-id: b61528c6c817127974acdc4635bccf607532287f
2021-04-13 02:52:56 -07:00
..
ddp_comm_hooks [Gradient Compression] Make orthogonalization_epsilon configurable in PowerSGDState (#55738) 2021-04-13 02:52:56 -07:00
__init__.py [Gradient Compression] Add unit tests that test default Python comm hook implementations (#47158) 2020-11-06 00:28:09 -08:00