mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Trevor Grant <trevor.d.gr...@gmail.com>
Subject Re: [VOTE] Apache Mahout 0.13.0 Release Candidate
Date Mon, 17 Apr 2017 00:44:59 GMT
+1 (binding)

Judicial Opinion:

Verified Signatures

Built with `mvn clean package -Phadoop2` ; `mvn clean package -Phadoop2
-Pviennacl-omp` ; `mvn clean package -Phadoop2 -DskipTests` (see below)
For all builds, ran `spark item similarity` example with no issue, `wiki`
example fails with non-mahout related file not found errors. need JIRA to
update file path- as this is an example where the functionality is non
related to Mahout, i find this non blocking bc it is likely all of the
prior releases are now also broken (that is, if we had released this a
month ago, the release would still now be broken)

For viennacl, the build fails on the `sparse mmul microbench test`.  I have
two machines, both fail.  Both have older graphics cards (specs below).  On
the older of the two, I am able to make something very similar to the test
pass in the shell if I set `s=625` (using `timeSparseDRMMMul`). This
implies that the functionality is sound, but my cards simply aren't new
enough to pass the test.  I recommend opening a JIRA to tune down the
aggressiveness of the test- e.g. making `s=500` (currently `s=1000`).

This recommendation is based on 1) my cards aren't _that_ old and 2) I
don't want to buy new graphics cards just to pass unit tests.
(Included CPU arch for reference wrt OMP)

Dev Box:

NVidia GeForce GT 740
Driver Version: 352.63
Memory: 1021MiB

ViennaCL v 1.7.0

CPU: Intel Core i7-3770K @3.50 Ghz (Ivy Bridge Micro Arch- 3rd Gen)

Ubuntu 14.04.3 LTS


Laptop:
NVidia GeForce GTX 960M
Driver Version: 367.57
Memory: 2002MiB

CPU: Intel Core i7-5500U @ 2.40 Ghz (Broadwell-U Micro arch - 5th Gen)

ViennaCL v 1.7.0

Ubuntu 16.04.1 LTS



Trevor Grant
Data Scientist
https://github.com/rawkintrevo
http://stackexchange.com/users/3002022/rawkintrevo
http://trevorgrant.org

*"Fortunate is he, who is able to know the causes of things."  -Virgil*


On Sat, Apr 15, 2017 at 11:35 PM, Andrew Palumbo <ap.dev@outlook.com> wrote:

> +1 (binding)
>
>
> Built and tested source distribution with both profiles -Pviennacl and
> -Pviennacl-omp.
>
>
> Ran SparseSparseDrmTimer.mscala through the shell in both pseudo cluster
> and local[2] mode
>
> Tested with several iterations and combinations of arguments eg:
>
> timeSparseDRMMMul(1000,1000,1000,5,.2,1234L)
>
> Mostly with out issue in a consumer grade card.
>
> Note: in the shell after the %*% is called by a partition and the GPU is
> in use, will get a native failure exception which is caught and allows for
> MMul of that partition to fall back to JVM MMul (single-threaded).. this
> should be changed in 0.13.1 to fall back to OpenMP MMul.
>
> Note: Binary distribution is built for Tesla GPUs. My card was not
> compatible, though out target is higher end GPUs on AWS or PowerPC (PowerPC
> uses  teslas) so not a blocker IMO.
>
> We will target a wider range of cards in the next distributions.
>
>
>
>
>
>
>
>
>
>
>
> ________________________________
> From: Andrew Musselman <andrew.musselman@gmail.com>
> Sent: Saturday, April 15, 2017 2:48:17 AM
> To: user@mahout.apache.org; dev@mahout.apache.org
> Subject: Re: [VOTE] Apache Mahout 0.13.0 Release Candidate
>
> Hashes and sigs confirmed, bin and src (viennacl and viennacl-omp)
> artifacts run the spark shell and the sparse drm test fine, and kick off
> the GPU.
>
> +1 (binding)
>
> On Fri, Apr 14, 2017 at 10:25 PM, Andrew Musselman <
> andrew.musselman@gmail.com> wrote:
>
> > This is the vote for release 0.13.0 of Apache Mahout.
> >
> > The vote will be going for at least 72 hours and will be closed on
> > Monday, April 17th, 2017 or once there are at least 3 PMC +1 binding
> votes
> >  (whichever occurs earlier).  Please download, test and vote with
> >
> > [ ] +1, accept RC as the official 0.13.0 release of Apache Mahout
> > [ ] +0, I don't care either way,
> > [ ] -1, do not accept RC as the official 0.13.0 release of Apache Mahout,
> > because...
> >
> >
> > Maven staging repo:
> >
> > *https://repository.apache.org/content/repositories/
> orgapachemahout-1044/org/apache/mahout/apache-mahout-distribution/0.13.0
> > <https://repository.apache.org/content/repositories/
> orgapachemahout-1044/org/apache/mahout/apache-mahout-distribution/0.13.0>*
> >
> > The git tag to be voted upon is mahout-0.13.0
> >
>

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