This repository has been archived by the owner on Nov 17, 2023. It is now read-only.
[v1.x] Update onnx support to work with onnx 1.7.0 with most CV models #19017
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
* fix pooling_convention warning * fix pooling_convention warning * fix lint Co-authored-by: JackieWu <[email protected]>
|
Hey @josephevans , Thanks for submitting the PR
CI supported jobs: [centos-cpu, windows-cpu, edge, miscellaneous, unix-cpu, sanity, website, unix-gpu, centos-gpu, clang, windows-gpu] Note: |
added 3 commits
August 26, 2020 16:36
josephevans
changed the title
added 15 commits
August 27, 2020 20:24
…rsion supported by onnx. Pass this parameter to each graph layer when exporting.
…which onnx opset version they want to use when exporting, defaults to latest version supported by onnx.
…n time, we know what opset version to use.
…bj, which reflects what opset version the onnx model uses.
…et versions. Check opset version instead of ONNX version for Pooling operator.
…rom gluon model zoo.
josephevans
requested review from
aaronmarkham and
marcoabreu
as code owners
…8675) Signed-off-by: Serge Panev <[email protected]>
…ions - this gets all ONNX unit tests to pass with onnx 1.7.
…ormal unittests pipelines.
josephevans
changed the title
josephevans
changed the title
josephevans
changed the title
vandanavk
reviewed
Sep 3, 2020
There was a problem hiding this comment.
Could update documentation as relevant
https://cwiki.apache.org/confluence/display/MXNET/ONNX
https://mxnet.apache.org/api/python/docs/tutorials/deploy/export/onnx.html#Prerequisites
|
@Kh4L for review please. Thank you. |
Kh4L
approved these changes
Sep 3, 2020
added 2 commits
September 8, 2020 21:18
… to install onnx and onnxruntime in the CentOS containers.
|
@vandanavk Refactored based on your review, could you please approve? thanks, |
vandanavk
approved these changes
Sep 11, 2020
|
@josephevans what's the status of adding the PR to the master branch? |
leezu
pushed a commit
to leezu/mxnet
that referenced
this pull request
Oct 8, 2020
Forward port of apache#19017 Co-authored-by: Joe Evans <[email protected]>
|
Thanks for bringing it up. It should be on 1.X and Joe had plans for the same. @josephevans |
josephevans
added a commit
to josephevans/mxnet
that referenced
this pull request
Nov 20, 2020
Forward port of apache#19017 Co-authored-by: Joe Evans <[email protected]>
josephevans
added a commit
to josephevans/mxnet
that referenced
this pull request
Nov 23, 2020
Forward port of apache#19017 Co-authored-by: Joe Evans <[email protected]>
7 tasks
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This PR improves MXNet support for onnx import and export for newer onnx versions (including the latest, 1.7.0).
The goal of this PR is to allow exporting of the most common CV models used and verify inference with the latest onnxruntime (1.4.0).
With newer onnx opset versions, the attributes for some operators have been changed to inputs, so most of the work in the two _op_translations.py files relates to converting between attributes and inputs. Updated operator translations between MXNet and onnx as follows:
Export operators:
Import operators:
This PR also changes the import and export code to pass an “opset_version” variable which defaults to the latest version supported by the installed onnx module. This variable is used to determine what format to use for onnx conversions.
This PR also cherry-picks the following PRs from the master branch:
This PR also updates the CI environments to install the latest version of onnx (1.7.0) and onnxruntime (1.4.0) for testing model export from mxnet and inference using onnxruntime.
All onnx unit tests are passing. An additional onnx test has been added in tests/python-pytest/onnx/test_onnxruntime.py which tests the following models (from gluoncv model zoo) for export and inference in onnxruntime:
There are still some operators that need to be updated to support the latest opset versions.