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This repository has been archived by the owner on Nov 17, 2023.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

[MXNET-1086] added sub and mul to ONNX->TensorRT conversion #15344

Merged
merged 2 commits into from
Jun 25, 2019
Merged

[MXNET-1086] added sub and mul to ONNX->TensorRT conversion #15344

merged 2 commits into from
Jun 25, 2019

Conversation

bermanmaxim
Copy link
Contributor

@bermanmaxim bermanmaxim commented Jun 24, 2019 •
edited
Loading

Description

Added the sub/mul operators to the MXnet -> ONNX conversion, for usage in the ONNX -> tensorRT conversion.

Checklist

Essentials

Please feel free to remove inapplicable items for your PR.

  • The PR title starts with [MXNET-$JIRA_ID], where $JIRA_ID refers to the relevant JIRA issue created (except PRs with tiny changes)
  • Changes are complete (i.e. I finished coding on this PR)
  • All changes have test coverage:
  • Unit tests are added for small changes to verify correctness (e.g. adding a new operator)
  • Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore)
  • Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL)
  • Code is well-documented:
  • For user-facing API changes, API doc string has been updated.
  • For new C++ functions in header files, their functionalities and arguments are documented.
  • For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable
  • Check the API doc at http://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$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

  • Added ConvertElementwiseSub in nnvm_to_onnx-inl.h / nnvm_to_onnx.cc
  • Added ConvertElementwiseMul in nnvm_to_onnx-inl.h / nnvm_to_onnx.cc
  • Minor: fixed typo in CMakelists

Comments

  • The implementation directly follows ConvertElementwiseAdd.
  • Fixes a bug since these missing operators were registered as TRTcompatible even before this commit.

@KellenSunderland
Copy link
Contributor

KellenSunderland commented Jun 24, 2019 •
edited
Loading

So I was going to point out that we'd need to update the subgraph selector to add support for the mul / sub but it seems like that's already there. This may actually fix model conversions for models that have those ops. LGTM, any thoughts from your side @Caenorst?

https://github.com/apache/incubator-mxnet/blob/master/src/operator/subgraph/tensorrt/tensorrt-inl.h#L93

@Caenorst
Copy link
Contributor

Yes, that was a mistake on my side.
The PR looks fine to me.

@KellenSunderland
Copy link
Contributor

Test looks good, thanks for adding it.

@KellenSunderland KellenSunderland merged commit 51acd4d into apache:master Jun 25, 2019
marcoabreu
marcoabreu reviewed Jun 26, 2019
View reviewed changes
tests/python/tensorrt/test_ops.py
mx.contrib.tensorrt.set_use_fp16(original_precision_value)


def test_elementwise():
Copy link
Contributor

Choose a reason for hiding this comment

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

with_seed missing

@KellenSunderland KellenSunderland mentioned this pull request Aug 13, 2019
Closed
KellenSunderland pushed a commit to KellenSunderland/incubator-mxnet that referenced this pull request Aug 13, 2019
@bermanmaxim @KellenSunderland
[MXNET-1086] added sub and mul to ONNX->TensorRT conversion (apache#1…
b0b14c5 
…5344)

* added sub and mul to ONNX->TensorRT conversion

* add test for elementwise ops in TRT
TaoLv pushed a commit that referenced this pull request Aug 16, 2019
@KellenSunderland @TaoLv
[MXNET-1086] added sub and mul to ONNX->TensorRT conversion (#15344) (#…
964f288 
…15875)

* added sub and mul to ONNX->TensorRT conversion

* add test for elementwise ops in TRT
access2rohit pushed a commit to access2rohit/incubator-mxnet that referenced this pull request Sep 6, 2019
Revert "[MXNET-1086] added sub and mul to ONNX->TensorRT conversion (a…
4e32d7d 
…pache#15344)"

This reverts commit 51acd4d.
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Reviewers

@marcoabreu marcoabreu marcoabreu left review comments

@rahul003 rahul003 Awaiting requested review from rahul003

@szha szha Awaiting requested review from szha szha is a code owner

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@bermanmaxim @KellenSunderland @Caenorst @marcoabreu