Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

[MXNET-1421] Added (CuDNN)BatchNorm operator to the list of mirrored operators #16022

Merged

Conversation

ArmageddonKnight
Copy link
Contributor

Description

Added the (CuDNN)BatchNorm operator to the list of mirrored operators.

Checklist

Essentials

Please feel free to remove inapplicable items for your PR.

  • The PR title starts with [MXNET-$JIRA_ID], where $JIRA_ID refers to the relevant JIRA issue created (except PRs with tiny changes)
  • Changes are complete (i.e. I finished coding on this PR)
  • Code is well-documented:
  • For user-facing API changes, API doc string has been updated.
  • For new C++ functions in header files, their functionalities and arguments are documented.
  • For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable
  • Check the API doc at http:https://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$PR_ID/$BUILD_ID/index.html
  • To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change

Changes

  • In file src/executor/graph_executor.cc method GraphExecutor::InitFullGraph, removed the restriction on the (CuDNN)BatchNorm operator.
  • In file src/operator/cudnn_batch_norm-inl.h class CuDNNBatchNormOp, added an internal lock on auxiliary states to prevent multiple consecutive updates on the moving mean and variance.

Comments

I have documented the changes here. Any comment is welcome. Thanks.

FYI, @antinucleon

S.A. Issue #14383

@karan6181
Copy link
Contributor

@mxnet-label-bot add [Operator, Cuda, pr-awaiting-review]

Copy link
Member

@wkcn wkcn left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good catch! LGTM. Thank you for the fix!

@eric-haibin-lin
Copy link
Member

@antinucleon

@eric-haibin-lin
Copy link
Member

@ArmageddonKnight can you look into the sync batch norm failure?

@ArmageddonKnight
Copy link
Contributor Author

@eric-haibin-lin I think I fixed the sync_batch_norm test failure.

Copy link
Member

@eric-haibin-lin eric-haibin-lin left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

thanks. let me trigger the CI again to see if it passes the last website test

@eric-haibin-lin
Copy link
Member

There was some update in mxnet website build. Could you sync with mxnet master? thanks

@eric-haibin-lin eric-haibin-lin merged commit 7d4f2f3 into apache:master Nov 13, 2019
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
CUDA Operator pr-awaiting-review PR is waiting for code review
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

5 participants