Wrap Caffe2 (RowWise)SparseAdagrad fusion operator as a PT op (#39904)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/39904
This diff wraps Caffe2's (RowWise)SparseAdagrad fusion operator on GPU as a PT op.
Reviewed By: jianyuh
Differential Revision: D22010193
fbshipit-source-id: 5df3c506c0dadd3b21180829fd2d5084ac76abc3