I wonder what is the current approach to reproduce random failures of
the tests battery?
would it be feasible to suggest (PR) for sklearn/test_setup.py (I guess)
to seed RNGs with some random but known seed and print it out (ideally if
nose's verbosity is exposed then only in verbose mode) so exactly the same run
could be easily reproduced later on happen failure occur.
My list of failures atm while sweeping through distributions:
1. .Segmentation fault
(unfortunately wasn't ran in verbose mode -- will do from now on)
32bit debian wheezy
2.
ubuntu 11.04 i386 as well
....................................
======================================================================
ERROR: sklearn.svm.tests.test_sparse.test_sparse_svc_clone_with_callable_kernel
----------------------------------------------------------------------
Traceback (most recent call last):
File "/usr/lib/pymodules/python2.7/nose/case.py", line 186, in runTest
self.test(*self.arg)
File
"/tmp/buildd/scikit-learn-0.11.0/debian/python-sklearn/usr/lib/python2.7/dist-packages/sklearn/svm/tests/test_sparse.py",
line 210, in test_sparse_svc_clone_with_callable_kernel
b.fit(X_sp, Y)
File
"/tmp/buildd/scikit-learn-0.11.0/debian/python-sklearn/usr/lib/python2.7/dist-packages/sklearn/svm/base.py",
line 141, in fit
fit(X, y, sample_weight)
File
"/tmp/buildd/scikit-learn-0.11.0/debian/python-sklearn/usr/lib/python2.7/dist-packages/sklearn/svm/base.py",
line 278, in _sparse_fit
int(self.shrinking), int(self.probability))
File "libsvm_sparse.pyx", line 158, in
sklearn.svm.libsvm_sparse.libsvm_sparse_train (sklearn/svm/libsvm_sparse.c:1927)
File "/usr/lib/python2.7/dist-packages/scipy/sparse/compressed.py", line 75,
in __init__
self.shape = shape # spmatrix will check for errors
File "/usr/lib/python2.7/dist-packages/scipy/sparse/base.py", line 71, in
set_shape
raise ValueError('invalid shape')
ValueError: invalid shape
----------------------------------------------------------------------
Ran 765 tests in 48.439s
FAILED (SKIP=12, errors=1)
3. I lost record of it but it was about ICA and failure to equate 1.0 to
0.994... up to two decimals which looked weird to me (do not remember
location or details -- sorry)
--
Yaroslav O. Halchenko
Postdoctoral Fellow, Department of Psychological and Brain Sciences
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
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
