Enable MKL-DNN in pip packages #16899

TaoLv · 2019-11-25T07:53:12Z

Description

Discussion: https://lists.apache.org/thread.html/1a22dbd79098adab6d02d16e8d607bae2acc908c0bb1b085d28a51ba@%3Cdev.mxnet.apache.org%3E

Checklist

Essentials

Please feel free to remove inapplicable items for your PR.

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

Feature1, tests, (and when applicable, API doc)
Feature2, tests, (and when applicable, API doc)

Comments

If this change is a backward incompatible change, why must this change be made.
Interesting edge cases to note here

TaoLv · 2019-11-25T08:40:46Z

@lanking520 @perdasilva

szha

Given the changes, it doesn't quite make sense to keep the *mkl variants anymore, since that's the only difference.

TaoLv · 2019-11-27T05:45:40Z

I'm not sure whether the backward compatibility matters for convenient binary releases or not. What about if downstream projects explicitly call pip install mxnet-cu90mkl in their scripts [1]?
@pengzhao-intel suggested to keep these *mkl variants till 2.0 major release [2].

[1]https://github.com/dmlc/gluon-cv/blob/master/tests/py3.yml#L19
[2]https://lists.apache.org/thread.html/87282a60cd73257d3dd4b57401ffe2eabd293d77456d9966b1f91336@%3Cdev.mxnet.apache.org%3E

pengzhao-intel · 2019-11-28T02:18:27Z

@szha does this make sense to keep *mkl variants for a short time? Or we need to remove from now.
@marcoabreu @samskalicky @larroy any comments?

I am going to merging the PR in the early of next week in case there is no objection.

…to release-mkldnn

szha · 2019-12-01T00:02:04Z

Not really. Given that the content would be the same, having both would just waste space.

wkcn · 2019-12-01T00:09:35Z

I think keeping *mkl is better, since users may use *mkl to install MXNet.
To save space, we can add a dummy package, whose dependency is MXNet package, and raise the deprecated warning.

pengzhao-intel · 2019-12-01T08:22:18Z

Not really. Given that the content would be the same, having both would just waste space.

@szha @wkcn Agree. We will try if dummy package works now.

…to release-mkldnn

TaoLv · 2019-12-04T02:44:10Z

For these .mk files, I git rm those non-mkl variants and then git mv those mkl variants to non-mkl.

…to release-mkldnn Conflicts: make/pip/pip_linux_mkl.mk make/pip/pip_linux_native.mk make/staticbuild/darwin_mkl.mk make/staticbuild/linux_cu100mkl.mk make/staticbuild/linux_cu101mkl.mk make/staticbuild/linux_cu75mkl.mk make/staticbuild/linux_cu80mkl.mk make/staticbuild/linux_cu90mkl.mk make/staticbuild/linux_cu91mkl.mk make/staticbuild/linux_cu92mkl.mk make/staticbuild/linux_mkl.mk

…to release-mkldnn

TaoLv · 2019-12-26T09:25:03Z

@szha @perdasilva Could you please review? Is it possible to verify these changes in dev CD? Thanks.

…to release-mkldnn

larroy · 2020-01-04T02:44:29Z

R-package/src/im2rec.h


 #include <Rcpp.h>
 #include <string>
+#include <opencv2/core/version.hpp>


Is this change related or is your branch out of date?

larroy · 2020-01-04T02:45:20Z

Can you rebase?

TaoLv · 2020-01-04T09:24:42Z

Can you rebase?

Done.

…to release-mkldnn Conflicts: cd/Jenkinsfile_cd_pipeline cd/Jenkinsfile_release_job cd/README.md

TaoLv · 2020-01-26T15:51:26Z

@szha @perdasilva @leezu @larroy @pengzhao-intel Could you please review? How can we proceed this?

leezu

LGTM

…to release-mkldnn Conflicts: tools/staticbuild/README.md

samskalicky · 2020-02-10T18:08:28Z

@TaoLv can you also update the cu102.mk files:
https://github.com/apache/incubator-mxnet/blob/master/make/staticbuild/linux_cu102.mk
https://github.com/apache/incubator-mxnet/blob/master/make/staticbuild/linux_cu102mkl.mk

TaoLv · 2020-02-11T01:48:12Z

Thank you @samskalicky. I over looked these files when merging the master branch. Will update them.

szha · 2020-02-11T01:51:13Z

tools/staticbuild/README.md


 For the CMake build, you need to install `patchelf` first, for example via `apt
 install patchelf` on Ubuntu systems.
+>>>>>>> ae145cdb1de95e8275dd087585f4aab69bc94005


seems to be a leftover from resolving merge conflict

Thanks. It's fixed.

samskalicky

LGTM, I restarted your windows GPU run since the cmake.exe error is still flaky. But im not sure about the website error:

[2020-02-11T02:14:04.984Z] Extension error:
[2020-02-11T02:14:04.984Z] Could not import extension autodocsumm (exception: cannot import name 'Signature' from 'sphinx.ext.autodoc' (unknown location))
[2020-02-11T02:14:05.247Z] make[1]: *** [html] Error 2
[2020-02-11T02:14:05.247Z] Makefile:95: recipe for target 'html' failed
[2020-02-11T02:14:05.247Z] make[1]: Leaving directory '/work/mxnet/docs/python_docs/python/build'

@aaronmarkham any idea?

TaoLv · 2020-02-11T06:30:00Z

Thank you @samskalicky . See #17560 for the website faliure.

…to release-mkldnn

TaoLv · 2020-02-14T05:33:08Z

@szha @leezu @samskalicky Is it good to go? Let me know if anything need be adjusted. Thanks.

leezu · 2020-02-14T21:22:51Z

@TaoLv Sheng just merged fixes for the CD pipeline. Let's wait 24 hours to verify CD works on master branch. Then we can merge this PR.

szha · 2020-02-16T18:58:11Z

Looks like the update has taken effect, though the jenkins jobs are still running for *mkl variants. http://jenkins.mxnet-ci.amazon-ml.com/blue/organizations/jenkins/restricted-mxnet-cd%2Fmxnet-cd-release-job/detail/mxnet-cd-release-job/724/pipeline

TaoLv · 2020-02-17T02:13:10Z

@szha What can I do for this situation? It seems we need change the value of MXNET_VARIANTS in the build environment.

* enable mkldnn by default in pip packages * add make file for native build * rm *mkl build scripts * remove *mkl variants * Change readme/cd * clear native build configurations * fix static build test in ci * git mv linux_cu102mkl.mk linux_cu102.mk * fix merge conflict

* Fix cd/mxnet_lib/dynamic/Jenkins_pipeline.groovy Fixes a regression in CD introduced by #17645 * Fix whitespace * Add NATIVE_ADDITIONAL.md Fixes a regression in CD introduced by #16899 * Update other $TYPE_ADDITIONAL.MD * Fix cd/python/docker Fixes regression introduced by #15990

* Fix cd/mxnet_lib/dynamic/Jenkins_pipeline.groovy Fixes a regression in CD introduced by apache#17645 * Fix whitespace * Add NATIVE_ADDITIONAL.md Fixes a regression in CD introduced by apache#16899 * Update other $TYPE_ADDITIONAL.MD * Fix cd/python/docker Fixes regression introduced by apache#15990

* enable mkldnn by default in pip packages * add make file for native build * rm *mkl build scripts * remove *mkl variants * Change readme/cd * clear native build configurations * fix static build test in ci * git mv linux_cu102mkl.mk linux_cu102.mk * fix merge conflict

* Fix cd/mxnet_lib/dynamic/Jenkins_pipeline.groovy Fixes a regression in CD introduced by apache#17645 * Fix whitespace * Add NATIVE_ADDITIONAL.md Fixes a regression in CD introduced by apache#16899 * Update other $TYPE_ADDITIONAL.MD * Fix cd/python/docker Fixes regression introduced by apache#15990

* Fix CD (#17776) * Fix cd/mxnet_lib/dynamic/Jenkins_pipeline.groovy Fixes a regression in CD introduced by #17645 * Fix whitespace * Add NATIVE_ADDITIONAL.md Fixes a regression in CD introduced by #16899 * Update other $TYPE_ADDITIONAL.MD * Fix cd/python/docker Fixes regression introduced by #15990 * [CD] update pypi description, setup.py (#17681) * update pypi description, setup.py, use manylinux2010, use unified dist link for nightly * Use manylinux2014 Co-authored-by: Leonard Lausen <leonard@lausen.nl> * reverting .so path as per MAKE flow Co-authored-by: Leonard Lausen <lausen@amazon.com> Co-authored-by: Sheng Zha <szha@users.noreply.github.com> Co-authored-by: Leonard Lausen <leonard@lausen.nl>

* Fix CD (apache#17776) * Fix cd/mxnet_lib/dynamic/Jenkins_pipeline.groovy Fixes a regression in CD introduced by apache#17645 * Fix whitespace * Add NATIVE_ADDITIONAL.md Fixes a regression in CD introduced by apache#16899 * Update other $TYPE_ADDITIONAL.MD * Fix cd/python/docker Fixes regression introduced by apache#15990 * [CD] update pypi description, setup.py (apache#17681) * update pypi description, setup.py, use manylinux2010, use unified dist link for nightly * Use manylinux2014 Co-authored-by: Leonard Lausen <leonard@lausen.nl> * reverting .so path as per MAKE flow Co-authored-by: Leonard Lausen <lausen@amazon.com> Co-authored-by: Sheng Zha <szha@users.noreply.github.com> Co-authored-by: Leonard Lausen <leonard@lausen.nl>

enable mkldnn by default in pip packages

8259349

TaoLv requested a review from szha as a code owner November 25, 2019 07:53

pengzhao-intel added the MKLDNN label Nov 25, 2019

szha reviewed Nov 27, 2019

View reviewed changes

TaoLv added 2 commits December 1, 2019 05:59

Merge branch 'master' of https://github.com/apache/incubator-mxnet in…

284aa18

…to release-mkldnn

add make file for native build

91bd30e

TaoLv added 2 commits December 4, 2019 10:23

Merge branch 'master' of https://github.com/apache/incubator-mxnet in…

39e1b5c

…to release-mkldnn

rm *mkl build scripts

2fb6e43

TaoLv added 3 commits December 17, 2019 10:44

remove *mkl variants

7195c78

Merge branch 'master' of https://github.com/apache/incubator-mxnet in…

9417588

…to release-mkldnn

TaoLv requested review from aaronmarkham and marcoabreu as code owners December 26, 2019 06:41

TaoLv changed the title ~~[WIP] Enable MKL-DNN in pip packages~~ Enable MKL-DNN in pip packages Dec 26, 2019

TaoLv added 4 commits December 26, 2019 19:05

Merge branch 'master' of https://github.com/apache/incubator-mxnet in…

4fe1eaa

…to release-mkldnn

Change readme/cd

912b921

clear native build configurations

79bb7b3

fix static build test in ci

055e3c0

larroy reviewed Jan 4, 2020

View reviewed changes

Merge branch 'master' of https://github.com/apache/incubator-mxnet in…

84ae814

…to release-mkldnn Conflicts: cd/Jenkinsfile_cd_pipeline cd/Jenkinsfile_release_job cd/README.md

leezu approved these changes Jan 26, 2020

View reviewed changes

Merge branch 'master' of https://github.com/apache/incubator-mxnet in…

3418636

…to release-mkldnn Conflicts: tools/staticbuild/README.md

szha reviewed Feb 11, 2020

View reviewed changes

samskalicky approved these changes Feb 11, 2020

View reviewed changes

TaoLv added 3 commits February 11, 2020 18:39

git mv linux_cu102mkl.mk linux_cu102.mk

d2aec9d

fix merge conflict

94cc6b0

Merge branch 'master' of https://github.com/apache/incubator-mxnet in…

8b67d21

…to release-mkldnn

leezu merged commit 2f6cdd3 into apache:master Feb 15, 2020

leezu mentioned this pull request Feb 25, 2020

mxnet=1.6.0 for OSX on pypi does not work #17671

Closed

leezu mentioned this pull request Mar 5, 2020

Fix CD #17776

Merged

leezu mentioned this pull request Mar 6, 2020

Fix CD #17784

Merged

jasperzhong mentioned this pull request Nov 22, 2020

MKLDNN header missing in mxnet-cu102==1.7.0 pip package #19575

Closed

Enable MKL-DNN in pip packages #16899

Enable MKL-DNN in pip packages #16899

Uh oh!

Conversation

TaoLv commented Nov 25, 2019

Description

Checklist

Essentials

Changes

Comments

Uh oh!

TaoLv commented Nov 25, 2019

Uh oh!

szha left a comment

Choose a reason for hiding this comment

Uh oh!

TaoLv commented Nov 27, 2019

Uh oh!

pengzhao-intel commented Nov 28, 2019

Uh oh!

szha commented Dec 1, 2019

Uh oh!

wkcn commented Dec 1, 2019

Uh oh!

pengzhao-intel commented Dec 1, 2019

Uh oh!

TaoLv commented Dec 4, 2019

Uh oh!

TaoLv commented Dec 26, 2019

Uh oh!

larroy Jan 4, 2020

Choose a reason for hiding this comment

Uh oh!

larroy commented Jan 4, 2020

Uh oh!

TaoLv commented Jan 4, 2020

Uh oh!

TaoLv commented Jan 26, 2020

Uh oh!

leezu left a comment

Choose a reason for hiding this comment

Uh oh!

samskalicky commented Feb 10, 2020

Uh oh!

TaoLv commented Feb 11, 2020

Uh oh!

szha Feb 11, 2020

Choose a reason for hiding this comment

Uh oh!

TaoLv Feb 11, 2020

Choose a reason for hiding this comment

Uh oh!

samskalicky left a comment

Choose a reason for hiding this comment

Uh oh!

TaoLv commented Feb 11, 2020

Uh oh!

TaoLv commented Feb 14, 2020

Uh oh!

leezu commented Feb 14, 2020

Uh oh!

szha commented Feb 16, 2020

Uh oh!

TaoLv commented Feb 17, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

TaoLv commented Feb 17, 2020 •

edited

Loading