pytorch
b05d8b95 - improve CPU performance for log_softmax when dim != -1 on both float32 and bfloat16 (#72163)

Commit

3 years ago

improve CPU performance for log_softmax when dim != -1 on both float32 and bfloat16 (#72163) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/72163 Pull Request resolved: https://github.com/pytorch/pytorch/pull/64726 Test Plan: Imported from OSS Reviewed By: VitalyFedyunin Differential Revision: D33862416 Pulled By: frank-wei fbshipit-source-id: 41359864348dc2425b22f7ae02883e95922192e2 (cherry picked from commit 377909dbe76f30806e4a933f638127e81470b4de)

References

#72894 - Merge pytorch master into lazy_tensor_staging

Author

mingfeima

Committer

pytorchmergebot

Parents

75e76944

Files3

aten/src/ATen/native
- SoftMax.cpp
- cpu
  - SoftMaxKernel.cpp
  - SoftmaxKernel.h

pytorch b05d8b95 - improve CPU performance for log_softmax when dim != -1 on both float32 and bfloat16 (#72163)

pytorch
b05d8b95 - improve CPU performance for log_softmax when dim != -1 on both float32 and bfloat16 (#72163)