On Mon, 08 Sep 2014, Olivier Grisel wrote:

> 2014-09-08 7:46 GMT-07:00 Yaroslav Halchenko <[email protected]>:

> > It is a bit early to say about Debian servers conclusively -- I have just
> > uploaded to Debian proper, so they have been rebuilt across
> > architectures:

> > https://buildd.debian.org/status/package.php?p=scikit-learn&suite=unstable
> > and armel seems to whine a bit:

> > ======================================================================
> > FAIL: sklearn.feature_extraction.tests.test_text.test_tfidf_no_smoothing
> > ----------------------------------------------------------------------
> > Traceback (most recent call last):
> >   File "/usr/lib/python2.7/dist-packages/nose/case.py", line 197, in runTest
> >     self.test(*self.arg)
> >   File 
> > "/«PKGBUILDDIR»/debian/python-sklearn/usr/lib/python2.7/dist-packages/sklearn/feature_extraction/tests/test_text.py",
> >  line 361, in test_tfidf_no_smoothing
> >     tr.fit_transform, X).toarray()
> >   File 
> > "/«PKGBUILDDIR»/debian/python-sklearn/usr/lib/python2.7/dist-packages/sklearn/utils/testing.py",
> >  line 211, in assert_warns_message
> >     % func.__name__)
> > AssertionError: No warning raised when calling fit_transform

> > ======================================================================
> > FAIL: sklearn.metrics.tests.test_metrics.test_precision_recall_curve_toydata
> > ----------------------------------------------------------------------
> > Traceback (most recent call last):
> >   File "/usr/lib/python2.7/dist-packages/nose/case.py", line 197, in runTest
> >     self.test(*self.arg)
> >   File 
> > "/«PKGBUILDDIR»/debian/python-sklearn/usr/lib/python2.7/dist-packages/sklearn/metrics/tests/test_metrics.py",
> >  line 1224, in test_precision_recall_curve_toydata
> >     assert_raises(Exception, precision_recall_curve, y_true, y_score)
> > AssertionError: Exception not raised

> > ----------------------------------------------------------------------
> > Ran 2753 tests in 1756.468s

> Interesting. It's great to have people test scikit-learn on the ARM
> platform. What is the armel architecture? how it compares to arm64 and
> armhf? 

I would better refer to e.g.
https://wiki.debian.org/ArmEabiPort
https://wiki.debian.org/Arm64Port
for the description besides stating that  armel is conformant to older
32bit ABI so it could run on all kinds of old ARM hardware (e.g. I have
some midnight gbox sitting under the TV running Debian within a chroot)

> Why the does arm64 build does not work for scikit-learn? Is
> there a missing dependency on that platform?

hm... actually not clear since it claims that it is because of missing
bdepends

scikit-learn build-depends on missing:
- libsvm-dev (>= 2.84.0)

while that one is available :-/  I will check

> Apparently those two errors on armel are caused by the fact that numpy
> does not raise an error when dividing by zero on this platform.

> I created an issue here:
> https://github.com/scikit-learn/scikit-learn/issues/3649

> It might be a bug in numpy or a fundamental limitation on armel. In
> the latter case we should update our code to take that into account.

> Do you know how to detect that we are running on armel from Python?
> e.g. what are the values of:

> from platform import processor, machine, architecture
> print(processor())
> print(machine())
> print(architecture())

I believe we did similar in pandas:

+        import platform
+        if platform.uname()[4].startswith('armv'):
+            raise nose.SkipTest("Fails on Debian arm boxes due to locales or 
whatelse")

so I guess I would need to do smth like that here as well, unless you
would do it upstream right away (after it gets cleared out, thanks for the
investigation above -- I forgot the detail if I knew it before)


-- 
Yaroslav O. Halchenko, Ph.D.
http://neuro.debian.net http://www.pymvpa.org http://www.fail2ban.org
Research Scientist,            Psychological and Brain Sciences Dept.
Dartmouth College, 419 Moore Hall, Hinman Box 6207, Hanover, NH 03755
Phone: +1 (603) 646-9834                       Fax: +1 (603) 646-1419
WWW:   http://www.linkedin.com/in/yarik        

------------------------------------------------------------------------------
Want excitement?
Manually upgrade your production database.
When you want reliability, choose Perforce
Perforce version control. Predictably reliable.
http://pubads.g.doubleclick.net/gampad/clk?id=157508191&iu=/4140/ostg.clktrk
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to