2012/5/9 Yaroslav Halchenko <[email protected]>: > > > On Wed, 09 May 2012, Olivier Grisel wrote: >> > so if it fails for some specific seed, I could check if it gets >> > replicated by running the same test again with the same seed. > >> > if it doesn't -- I know **for sure** that it is not related to having >> > random data but smth more fun, worth valgrinding for decisions based on >> > uninitialized memory etc. e.g. now it halted (burns cpu, doesn't return) >> > with >> > seed 1072 (before actually it just crashed in the same >> > sklearn.svm.tests.test_sparse.test_sparse_svc_clone_with_callable_kernel >> > and >> > didn't reproduce). > >> I am not sure I understand. You are saying that this test >> sklearn.svm.tests.test_sparse.test_sparse_svc_clone_with_callable_kernel >> crashes deterministically with seed 1072 for numpy.random while >> passing most of the time otherwise? > > > NO -- I am saying that > > a. it halted with SEED=1072 once -- I killed it > b. upon rerunning it with SEED=1072 it completed FINE
Alright so this is a nasty, data-independent, wrapper level bug. Thanks for the report. The valgrind report is probably the place to start. The randomized tests with environment variable-based seed control might still be useful to debug future failures related to python level RNG seeding though, as discussed on https://github.com/scikit-learn/scikit-learn/pull/842 . -- Olivier http://twitter.com/ogrisel - http://github.com/ogrisel ------------------------------------------------------------------------------ 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
