hm... I guess I would need to have a look at it -- there was a reason for using
old interface on how we treated returned bins... THANKS! -- we haven't got to
the 1.5.1 (although it is available... should upgrade ;))
as for running unittests -- if only "do not fail for sure" ones to be
ran, do
make unittest-nonlabile
because otherwise some tests are "ok" to fail from time to time to
effects of random data, e.g. like this one
Traceback (most recent call last):
File
"/opt/local/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/site-packages/mvpa/tests/test_datameasure.py",
line 112, in testAnalyzerWithSplitClassifier
len(self.dataset.uniquelabels)))
AssertionError:
Single scenario lead to failures of unittest testAnalyzerWithSplitClassifier:
on
clf=FeatureSelectionClassifier(clf=SVM(kernel_type='linear',
svm_impl='C_SVC', probability=1), enable_states=['training_confusion'],
descr='LinSVM on 5%(ANOVA)') :
We must have trained on each one more or less correctly. Got 66.666667%
correct on 2 labels
clf=FeatureSelectionClassifier(clf=SVM(kernel_type='linear',
svm_impl='C_SVC', probability=1), enable_states=['training_confusion'],
descr='LinSVM on 5%(SVM)') :
We must have trained on each one more or less correctly. Got 66.666667%
correct on 2 labels
On Sat, 29 Jan 2011, James Kyle wrote:
> Ok, here's a patch that fixes the first failed test aside from the lxml dep
> missing.
> On line #253 of mvpa/misc/transformers.py there's a version test for numpy
> that evaluates to True and creates a **kwargs hash of {'new': False} for
> N.histogram(). The 'new' kwarg is no longer supported as of Numpy version >=
> 1.4. I modified the test to provide a **kwarg of {} for all versions < 1.4
> resolving the test error.
> -james
> On Jan 29, 2011, at 11:00 AM, James Kyle wrote:
> > Ok, I rolled back to the 0.4 HEAD (maint/0.4).
> > This compiles without error, but running the unittests produces some
> > errors. The output of the log was quite large (2.5M) so rather than
> > attaching I uploaded it to our server.
> > The full log can be found here:
> > http://www.kspacemri.com/data/tests.log
> > -james
> > On Jan 28, 2011, at 5:43 PM, Yaroslav Halchenko wrote:
> >> On Fri, 28 Jan 2011, James Kyle wrote:
> >>>> ah... I guess it is time to release 0.4.6 of pymvpa -- we have that
> >>>> libsvm3.0 compatibility fix past 0.4.5 release:
> >>> I've actually been toying with the 0.6.0.dev branch, is that not ready
> >>> for the masses yet?
> >> well... it has been used by users and we hope to release it soon (
> >> probably as soon as our extensive testing battery shows no bad symptomps
> >> ;-) )
> >> the problem is that it is not compatible, so it might come as a surprise
> >> to macport users... for Debian users we ship python-mvpa-snapshot which
> >> provides 0.6.x functionality before we release it
> >> so altogether it would be better to stick with 0.4.x series until 0.6.0
> >> gets released
> >>>> or could macports simply build off a specific git branch? (i.e.
> >>>> maint/0.4)
> >>> I can setup the macports build script for whichever branch is
> >>> preferred/recommended.
> >> So you would not need to tune it up later on, probably let it stay as
> >> is -- HEAD of that repository corresponds to debian releases... I
> >> progressed it forward (merged maint/0.4)...
> >> is port taking a HEAD or releases (tags), as given by
> >> livecheck.regex {upstream\/([0-9.]+)}
> >> (and that is how it assigns the version according to (...) above)?
> >> --
> >> =------------------------------------------------------------------=
> >> Keep in touch www.onerussian.com
> >> Yaroslav Halchenko www.ohloh.net/accounts/yarikoptic
> >> _______________________________________________
> >> Pkg-ExpPsy-PyMVPA mailing list
> >> [email protected]
> >> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
> _______________________________________________
> Pkg-ExpPsy-PyMVPA mailing list
> [email protected]
> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
--
=------------------------------------------------------------------=
Keep in touch www.onerussian.com
Yaroslav Halchenko www.ohloh.net/accounts/yarikoptic
_______________________________________________
Pkg-ExpPsy-PyMVPA mailing list
[email protected]
http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
