systemml-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Nakul Jindal <naku...@gmail.com>
Subject Re: [VOTE] Apache SystemML 0.14.0-incubating (RC4)
Date Mon, 01 May 2017 05:27:29 GMT
+1

Ran the performance suite on an IBM internal cluster. Most tests seem to
perform reasonably close to the previous release.

-Nakul


On Fri, Apr 28, 2017 at 5:51 PM, Matthias Boehm <mboehm7@googlemail.com>
wrote:

> this regression is certainly something to look into but this release
> contains a large number of fixes including many that addressed severe OOM
> issues, so it might in fact be just an issue of more conservative but now
> correct execution plans given the current capabilities of our compiler.
>
> Regards,
> Matthias
>
> On Fri, Apr 28, 2017 at 5:39 PM, <dusenberrymw@gmail.com> wrote:
>
> > +1  Grabbed the tar binary and the tar source and tested various local
> > scripts in Scala & Python 2 + 3, and those ran fine.  However, I did run
> > the MNIST LeNet demo on both our 0.13 release and this 0.14 candidate,
> and
> > I noticed a regression in 0.14.  For the same script run back to back,
> the
> > 0.14 candidate took longer, and looking into the stats, on 0.13 there
> were
> > 864 Spark instructions executed, while on this 0.14 there were 2513 Spark
> > instructions executed.   This also brought the `sp_mapmm` and `sp_sel+`
> > instructions into the top 10 heavy hitters.  This could be related to the
> > issue that I am seeing in SYSTEMML-1561.
> >
> > Regardless, I'm still fine with releasing this, since the deep learning
> > support is still experimental for 0.14.  For our upcoming 1.0 release,
> all
> > engine bugs and issues related to deep learning need to be fixed.  Most
> of
> > these bugs are generally applicable to all algorithms, so it is in the
> > benefit of the project to fix them.
> >
> > --
> >
> > Mike Dusenberry
> > GitHub: github.com/dusenberrymw
> > LinkedIn: linkedin.com/in/mikedusenberry
> >
> > Sent from my iPhone.
> >
> >
> > > On Apr 28, 2017, at 10:37 AM, Arvind Surve <acs_s@yahoo.com.INVALID>
> > wrote:
> > >
> > > +1
> > > Completed following verifications   - License and Notice validations
>  -
> > Binary runtime validations    - Source code compilation and runtime
> > validations   - Python scripts validations using Python 2 Arvind Surve |
> > Spark Technology Center  | http://www.spark.tc/
> > >
> > >      From: Glenn Weidner <gweidner@us.ibm.com>
> > > To: dev@systemml.incubator.apache.org
> > > Sent: Monday, April 24, 2017 9:30 PM
> > > Subject: Re: [VOTE] Apache SystemML 0.14.0-incubating (RC4)
> > >
> > > +1
> > >
> > > Successfully ran Linear Regression, Logistic Regression, Naive Bayes,
> > SVM in
> > > Python notebooks with Spark 2.0.2 (in cloud environment) and Spark 2.1
> > (on local test cluster) after pip install of RC4 python artifact
> > > systemml-0.14.0-incubating-python.tgz. Also ran Linear Regression
> > Conjugate Gradient in Scala notebooks.
> > >
> > > Regards,
> > > Glenn
> > >
> > > Matthias Boehm ---04/24/2017 02:02:12 AM---+1 I ran large-scale
> > experiments on Spark 2.1 for L2SVM, GLM, MLogreg,
> > >
> > > From: Matthias Boehm <mboehm7@googlemail.com>
> > > To: dev@systemml.incubator.apache.org
> > > Date: 04/24/2017 02:02 AM
> > > Subject: Re: [VOTE] Apache SystemML 0.14.0-incubating (RC4)
> > >
> > >
> > >
> > > +1
> > >
> > > I ran large-scale experiments on Spark 2.1 for L2SVM, GLM, MLogreg,
> > > LinregCG, LinregDS, and PCA over scaled versions of MNIST and ImageNet
> > (up
> > > to 1TB, with uncompressed and compressed linear algebra) without any
> > > issues.
> > >
> > > Compared to previous experiments with SystemML 0.11 and Spark 1.6, I've
> > > seen substantial performance improvements of >2x for iterative
> algorithms
> > > with RDD operations in the inner loop over out-of-core datasets.
> > >
> > > Regards,
> > > Matthias
> > >
> > > On Wed, Apr 19, 2017 at 4:17 PM, Arvind Surve <acs_s@yahoo.com.invalid
> >
> > > wrote:
> > >
> > >> Please vote on releasing the following candidate as Apache SystemML
> > >> version 0.14.0-incubating !
> > >> The vote is open for at least 72 hours and passes if a majority of at
> > >> least 3 +1 PMC votes are cast.
> > >> [ ] +1 Release this package as Apache SystemML 0.14.0-incubating[ ] -1
> > Do
> > >> not release this package because ...
> > >> To learn more about Apache SystemML, please see
> http://systemml.apache.
> > >> org/
> > >> The tag to be voted on is v0.14.0-incubating-rc4 (
> > >> 8bdcf106ca9bd04c0f68924ad5827eb7d7d54952)
> > >> https://github.com/apache/incubator-systemml/commit/
> > >> 8bdcf106ca9bd04c0f68924ad5827eb7d7d54952
> > >>
> > >> The release artifacts can be found at :https://dist.apache.org/
> > >> repos/dist/dev/incubator/systemml/0.14.0-incubating-rc4/
> > >> The maven release artifacts, including signatures, digests, etc. can
> > >> be found at:https://repository.apache.org/content/repositories/
> > >> orgapachesystemml-1021/org/apache/systemml/systemml/0.14.
> 0-incubating/
> > >> ======================================= Apache Incubator release
> policy
> > >> =======================================Please find below the guide to
> > >> release management during incubation:http://incubator.
> > apache.org/guides/
> > >> releasemanagement.html
> > >> ========================================= How can I help test this
> > >> release? =========================================If you are a
> SystemML
> > >> user, you can help us test this release by taking an existing
> Algorithm
> > or
> > >> workload and running on this release candidate, thenreporting any
> > >> regressions.
> > >> ================================================== What justifies a
> -1
> > >> vote for this release? ==============================
> > ====================-1
> > >> votes should only occur for significant stop-ship bugs or legal
> > >> related issues (e.g. wrong license, missing header files, etc). Minor
> > bugs
> > >> or regressions should not block this release.
> > >>  -Arvind
> > >>  Arvind Surve | Spark Technology Center  | http://www.spark.tc/
> > >
> > >
> > >
> > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message