IRC logs of #shogun for Thursday, 2013-03-14

« Wednesday, 2013-03-13 Index Friday, 2013-03-15 »
--- Log opened Thu Mar 14 00:00:12 2013
-!- FSCV_ [~FSCV@65.19.131.247] has quit [Ping timeout: 245 seconds]00:13
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]00:21
-!- FSCV_ [~FSCV@vpnreactor-static44.den.icastcenter.com] has joined #shogun00:22
-!- FSCV [~FSCV@vpnreactor-static44.den.icastcenter.com] has joined #shogun01:15
-!- FSCV_ [~FSCV@vpnreactor-static44.den.icastcenter.com] has quit [Read error: Connection reset by peer]01:16
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]01:21
-!- FSCV [~FSCV@vpnreactor-static44.den.icastcenter.com] has quit [Quit: Leaving]01:54
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun06:32
-!- syst3mw0rm [~quassel@aws.aamirkhan.co.in] has quit [Remote host closed the connection]06:33
n4nd0blackburn: hey, did you see the results?06:35
-!- blackburn [~blackburn@83.234.54.160] has quit [Quit: Leaving.]08:32
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun09:01
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]09:50
wikingm000010:38
blackburnwiking: that's informative!10:40
wikingapt-get moo10:41
wiking:)10:41
wikingbtw is it only me or anybody else has MulticlassLibLinearTest.train segfaulting in unit tests?10:41
blackburnwiking: I had it but it disappeared after recompile10:42
wikingblackburn: recompile of...?10:42
blackburnwiking: tests10:42
wikingtrying10:42
wikingok that indeed helped10:43
blackburnwiking: do all tests run ok on your machine?10:50
wikingnow yes10:52
wikingi just had to make clean && make in tests10:52
-!- heiko1 [~heiko@nat-167-251.internal.eduroam.ucl.ac.uk] has joined #shogun11:14
heiko1sonney2k:  doing some final fixes now, let me know when you start11:14
heiko1blackburn:  around=?11:23
heiko1blackburn:  you should use EXPECT instead of ASSERT in the unit tests11:28
heiko1otherwise they stop11:28
heiko1but you want to continue checking to see whether the others worked11:28
heiko1assert only if program cannot continue, like division by zero11:28
heiko1blackburn:  multiclassocas segfault in unit test11:29
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]11:29
heiko1blackburn, ocas works again, dont know what caused the error11:56
wikingheiko1: sent11:56
heiko1 wiking :)11:57
heiko1sorry, I know too many Vi[c|k|ck]tors :)11:57
wiking;P11:57
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun12:08
shogun-notifier-shogun: Heiko Strathmann :master * 04e7e66 / examples/undocumented/python_modular/evaluation_clustering_simple.py: https://github.com/shogun-toolbox/shogun/commit/04e7e669b85ac4e080d1ce0159f9f648bad234c512:08
shogun-notifier-shogun: added new example for clustering evamluation which tests the method on simple toy data in order to decide whether it is good12:08
shogun-notifier-shogun: Heiko Strathmann :master * de3862a / examples/undocumented/python_modular/evaluation_clustering_simple.py: https://github.com/shogun-toolbox/shogun/commit/de3862a442ed1e5b52398c61f70b722ccc4e2b3f12:08
shogun-notifier-shogun: some minor changes in order to make integration test work12:08
shogun-notifier-shogun: Sergey Lisitsyn :master * 56cea1c / src/shogun/lib/tapkee/ (25 files): https://github.com/shogun-toolbox/shogun/commit/56cea1c5885752511157d327b15c871d0cd7aa5812:08
shogun-notifier-shogun: A big update for Tapkee12:08
shogun-notifier-shogun:12:08
shogun-notifier-shogun: - Fixed a few VS 2012 compilations12:08
shogun-notifier-shogun: - Parallel MDS12:08
shogun-notifier-shogun: - Parallel12:08
shogun-notifier-shogun: Sergey Lisitsyn :master * 934d00f / src/shogun/lib/tapkee/tapkee_shogun.cpp: https://github.com/shogun-toolbox/shogun/commit/934d00f78e8c4048c38803eb7afde9b2258fa04312:08
shogun-notifier-shogun: Fixed compilation12:08
shogun-notifier-shogun: Sergey Lisitsyn :master * 5fc7151 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/5fc7151692c81da1e27dcd137f7f2f4406986df412:08
shogun-notifier-shogun: Some improvement for multiclass confidences12:08
shogun-buildbotbuild #775 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/775  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>12:19
shogun-buildbotbuild #776 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/776  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com>12:24
shogun-notifier-shogun: Heiko Strathmann :master * ea53122 / src/shogun/labels/MulticlassLabels.cpp,src/shogun/labels/MulticlassLabels.h: https://github.com/shogun-toolbox/shogun/commit/ea53122fa195a6da3b916879ae26ec4bd3e85cbf12:26
shogun-notifier-shogun: added serialisation support for multiclass labels now that the confidences are stored in a matrix12:26
shogun-notifier-shogun: Heiko Strathmann :master * 68e1b97 / examples/undocumented/python_modular/evaluation_clustering.py: https://github.com/shogun-toolbox/shogun/commit/68e1b979b9f6e834cb7976f8c3165438e2c71c7c12:26
shogun-notifier-shogun: made example integration test work by storing everything including the now serialisable multiclass labels12:26
shogun-notifier-shogun: Heiko Strathmann :master * e7fd05f / / (3 files): https://github.com/shogun-toolbox/shogun/commit/e7fd05f9333343b56d73a9dc0e7691cee0af0c1112:26
shogun-notifier-shogun: Merge pull request #921 from karlnapf/master12:26
shogun-notifier-shogun:12:26
shogun-notifier-shogun: serialisation support for multiclass labels12:26
shogun-buildbotbuild #777 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/777  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>12:30
heiko1sonney2k: I changed a serialisation thing, we have to change the paramer version to 212:31
heiko1currently trying to do the mapping12:32
heiko1sonney2k: argh migration bugs :(12:39
heiko1sonney2k: Strange, if I leave the parameter version at 1 and just add the new multiclass label stuff to the map, it works12:42
heiko1so this way we can get at least the testsuite working12:42
heiko1while people wont be able to load serialised objects from before ....12:42
heiko1but this is something to fix later, what do you think?12:42
shogun-buildbotbuild #966 of deb1 - libshogun is complete: Failure [failed git]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/966  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>13:12
-!- votjak [~votjak@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun13:19
-!- votjak [~votjak@host-46-241-3-209.bbcustomer.zsttk.net] has left #shogun []13:21
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun14:08
heiko1sonney2k: serialization of matrices is broken ..... annoying14:14
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun14:15
n4nd0blackburn: hey there14:15
heiko1sonney2k:  there are still some debug messages not showing up14:25
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]14:45
wikingheiko1: here?14:48
heiko1wiking: use any example for serlialisation14:48
heiko1and enable debug14:48
wikingheiko1: so let's make it clear: what you would like to do basically14:49
heiko1not all the messages of Parameter.cpp are shown up14:49
wikingis that u can valgrind each and every unit test separately?14:49
heiko1wiking: or did you mean something else?14:49
heiko1wiking:14:49
heiko1yes14:49
heiko1to detect which one leaks14:49
wikingheiko1: should i compile it with --enable-malloc-trace?14:49
heiko1wiking: no14:49
heiko1wiking:  the traces are completely broken for unit tests14:49
heiko1just pure valgrind14:49
wikingok14:49
wikingchecking14:49
wikingjust rebased my repo14:50
wikingit'll take about 2 mins14:50
-!- blackburn [~lisitsin@mxs.kg.ru] has joined #shogun14:50
wikingheiko1: well what i could do is by default is a basic shellscript hack14:51
wikingheiko1: list all the modules14:51
blackburnheiko1: back14:51
wikingand then execute for each unit test module valgridn14:51
wikingheiko1: would that satisfy you?14:51
heiko1yes14:52
heiko1wiking: that would be good14:52
heiko1blackburn: fixing multiclass labels14:52
heiko1serilaization fails due to some uninitialized mem14:52
wikingheiko1: ok i'll create that shellscript hack into makefile... if i can ... if not then i'll add that shellscript14:53
heiko1wiking: either way, woudl be a great help14:53
wikingheiko1: btw have u checked whether there's a unit test malloc thingy with gtest?14:53
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]14:53
heiko1wiking: no14:53
heiko1 :)14:53
heiko1wiking: just letting you know what I would like to have :D14:54
wikingheiko1: heheh ok14:54
wikingmaybe there's a smoother way14:54
wikingi'll google a bit14:54
wikingbut i'll check now the valgrind output14:54
wikingafaik gperftools is a possible way to go ;P14:55
wikingbut imho it's the same as valgrind esentially14:55
wikingSparseFeaturesTest.serialization14:55
wiking?14:55
wikingthis?14:55
blackburnheiko1: what is wrong with uninited stuff?14:56
heiko1blackburn: ?14:56
blackburnheiko1: in multiclass labels14:56
heiko1blackburn: what do you mean?14:57
heiko1blackburn:  ah14:57
heiko1blackburn: serialization fails14:57
heiko1(and other things)14:57
heiko1investigating, but should be  fixed soon14:57
wikingheiko1: definitely lost: 10,257 bytes in 17 blocks14:58
wikingheiko1: doesn that sound right?14:58
heiko1wiking: in the whole tests?14:58
wikingheiko1: only this SparseFeaturesTest.serialization14:58
heiko1wiking: how did you find out?14:59
wikingwhat do you mean?14:59
wikingDYLD_LIBRARY_PATH=/Users/wiking/shogun/src/shogun:/Users/wiking/mosek/6/tools/platform/osx64x86/bin valgrind --leak-check=full ./shogun-unit-test --gtest_filter=SparseFeaturesTest.serialization14:59
wikingthis is what i did for execution14:59
wiking;)14:59
blackburnwiking: def lost is not normal you know :)14:59
heiko1wiking14:59
heiko1well14:59
heiko1thats what I want to have automagically :)14:59
wikingheiko1: i just wanted to check15:00
wikingbut yeah i get what u want ;)15:00
heiko1I see , cool15:00
heiko1yeah so we got the bastard ;)15:00
wikingbut i thought you've checked yourself15:00
wikingi'll pastebin this15:00
wikingbut unfortunately valgrind isn't so stable on 10.8 osx15:00
wiking:(15:00
wikinghttp://pastebin.com/qj8uPqmc15:02
wikingcheck this15:02
wikingin the meantime i'm creating the thingy ;)15:02
wikingheiko1: ok question15:08
wikingheiko1: what should be the commmand like? :)15:08
blackburnis that normal I don't understand what you are talking about? :)15:09
wikinghehehe15:09
wikingblackburn: heiko1 wants to be able to run automagically on each unit test module valgrind15:09
heiko1blackburn:  btw you should not use assertions in the tests, better expect, otherwise code will stop and we dont want it to stop but to check everything15:09
wikingbut separately15:10
heiko1wiking: 1sec15:10
wikingheiko1: i have now locally something like this15:10
wikingmaster unit] $ make valg                                                                                                                     [wiking@we02c096:~/shogun/tests/unit]15:10
wikingvalgrind SVMOcasTest.*15:10
wikingvalgrind IsomapTest.*15:10
wikingvalgrind MultidimensionaScalingTest.*15:10
wikingvalgrind CombinedFeaturesTest.*15:10
wikingvalgrind GaussianBlobsDataGenerator.*15:10
wikingvalgrind MeanShiftDataGenerator.*15:10
wikingvalgrind DenseFeaturesTest.*15:10
wikingvalgrind SparseFeaturesTest.*15:10
wikingvalgrind StringFeaturesTest.*15:10
wikingvalgrind CombinedKernelTest.*15:10
wikingvalgrind CustomKernelTest.*15:11
wikingbut it's echoed15:11
wikingso it's not yet executed15:11
wikingjust tell me what should be the name of the make command ;)15:11
wikingso but i guess you want something like this15:11
heiko1wiking: yeah15:13
heiko1make valgrind15:13
heiko1and then it runs all the tests15:13
wikingheiko1: that's already taken15:13
heiko1would be even better if it would run the inner ones also seperately15:13
wikingheiko1: make valgrind runs on the whole thing15:13
wikingthis one would run valgrind sperately on each 'module'15:14
wikingso <module>.* as i said15:14
wikingok i'll pr it and you'll see what i'm talking about15:16
heiko1wiking: maybe both possible?15:17
heiko1wiking: but one by one would be best15:17
wikingmake valgrind-per-module15:18
wikingi'll add one liek this15:18
wikingheiko1: https://github.com/shogun-toolbox/shogun/pull/92315:19
heiko1wiking: thanks!15:20
wikingshould i merge?15:20
shogun-notifier-shogun: Viktor Gal :master * b03de59 / tests/unit/Makefile: https://github.com/shogun-toolbox/shogun/commit/b03de59218b5261f86516e8689bdcb7c7d395bbf15:20
heiko1wiking:  :)15:20
shogun-notifier-shogun: Add new make option for valgrind15:20
shogun-notifier-shogun: 'make valgrind-per-module' command runs valgrind on each15:20
shogun-notifier-shogun: unit testing module separately.15:20
shogun-notifier-shogun: (for Heiko with love ;)15:20
shogun-notifier-shogun: Heiko Strathmann :master * ec0b158 / tests/unit/Makefile: https://github.com/shogun-toolbox/shogun/commit/ec0b1583f764005542ec1e9efedacab8f3f4273915:20
shogun-notifier-shogun: Merge pull request #923 from vigsterkr/master15:20
shogun-notifier-shogun:15:20
shogun-notifier-shogun: Add new make option for valgrind15:20
wikinghehe u were faster15:20
wiking;)15:20
heiko1haha15:20
heiko1sorry :)15:20
wikingnw15:20
wikingok so check it now15:20
heiko1I will15:21
wikingmake valgrind-per-module15:21
wikingand it'll run all of it15:21
wikingbut separately15:21
wikingi dont know how is it for you15:21
wikingbut for me each valgrind has leaks15:21
heiko1wiking:  gotta finish my stuff before. otherwise mess, but I will do soon!15:21
heiko1wiking: really?15:21
wikingbut yeah i guess it's because of valgrind being unstable on 10.815:21
heiko1ok15:21
wikingi'll check on another machine15:22
wikingwhere valgrind is good15:22
shogun-buildbotbuild #967 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/96715:30
shogun-buildbotbuild #778 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/778  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>15:31
shogun-buildbotbuild #779 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/779  blamelist: Viktor Gal <viktor.gal@maeth.com>15:36
wikingeh?15:37
wikingwhyyyyy15:37
wikingah ok this isn't my15:38
blackburnheiko1: ok will fix15:39
wiking:>>>15:40
shogun-notifier-shogun: Sergey Lisitsyn :master * 041ac8b / tests/unit/labels/MulticlassLabels_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/041ac8bfd4afc87ea3609c9a9e842c3f944f5cd315:40
shogun-notifier-shogun: Update MulticlassLabels_unittest.cc15:40
wikingah shit we need a fix15:42
-!- FSCV [~FSCV@65.19.131.247] has joined #shogun15:49
wikingheiko1: seems that actually CombinedKernelTest. has a problem16:00
heiko1wiking: uh16:00
wikingCombinedKernelTest.serialization to be exact16:00
heiko1I am haveing problems with serialization right now16:00
heiko1yeah16:00
wikingheiko1: i think that command needs polishing i've just commited16:00
wikingi'll fix it now16:00
heiko1maybe I will remove the serialization for multiclass labels for now16:00
wikingsince it doesn't work on ubuntu16:00
heiko1just give me a few more minutes to check16:01
blackburnheiko1: is it all ok with ocas?16:02
heiko1blackburn: not here, but I am currently deep in this other shit ...16:02
blackburnheiko1: oh so you are working as a plumber :)16:05
-!- FSCV_ [~FSCV@216-230-229-167-colo.oplink.net] has joined #shogun16:09
wikingIsomapTest.distance_preserving_max_k leaks16:11
wikingMultidimensionaScalingTest.distance_preserving16:11
wikingCustomKernelTest. leaks16:12
-!- FSCV [~FSCV@65.19.131.247] has quit [Ping timeout: 256 seconds]16:12
shogun-notifier-shogun: Viktor Gal :master * 4491057 / tests/unit/Makefile: https://github.com/shogun-toolbox/shogun/commit/449105724c4ef8d2c33e97786b27e4415769d0e516:15
shogun-notifier-shogun: Fix for previous commit.16:15
shogun-notifier-shogun: Left out the dependency for valgrind-per-module command16:15
shogun-notifier-shogun: Viktor Gal :master * 55d0cf0 / tests/unit/Makefile: https://github.com/shogun-toolbox/shogun/commit/55d0cf089274c8337eda912f8bf922ef46f6db4916:15
shogun-notifier-shogun: Add LD_LIBRARY_PATH for xargs in valgrind-per-module16:15
shogun-notifier-shogun: Viktor Gal :master * 83c5966 / tests/unit/Makefile: https://github.com/shogun-toolbox/shogun/commit/83c5966c3096ccb328c028efed78c9dbe3e957d316:15
shogun-notifier-shogun: Merge pull request #924 from vigsterkr/master16:15
shogun-notifier-shogun:16:15
shogun-notifier-shogun: Fix for valgrind-per-module make command16:15
-!- SCV__ [~FSCV@187.210.54.166] has joined #shogun16:17
-!- FSCV_ [~FSCV@216-230-229-167-colo.oplink.net] has quit [Ping timeout: 245 seconds]16:19
wikingheiko1: are you working on esting::TestCase::Run() (16:20
wikingfuck no16:20
heiko1?16:20
wikingso on this one: shogun::CGaussian::sample16:20
wiking?16:20
heiko1no16:20
heiko1go aheas16:20
heiko1d16:20
wikingok i'll fix it then16:21
wikingwho wrote this: MultidimensionaScalingTest_distance_preserving_Test16:21
wikingok should i add leaking issues?16:22
wikingi'll do16:22
wikingjust to use github a bit16:22
blackburnwiking: could you please paste lines?16:25
shogun-buildbotbuild #780 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/780  blamelist: Viktor Gal <vigsterkr@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com>, Viktor Gal <viktor.gal@maeth.com>16:25
blackburnof MultidimensionalScaling.cpp:***16:25
wikingblackburn: i'm just creating them16:25
wikingjust a sec16:25
blackburnokay thanks16:26
wikingok i've added some16:27
-!- FSCV_ [~FSCV@108.60.131.13] has joined #shogun16:32
-!- SCV__ [~FSCV@187.210.54.166] has quit [Ping timeout: 258 seconds]16:35
wikingok16:39
wikingi think i've put up all the leak reports16:39
wikinghave fun everybody16:39
-!- FSCV_ [~FSCV@108.60.131.13] has quit [Ping timeout: 260 seconds]16:41
blackburnwiking: thanks will take a look tonight16:45
wikingnw16:45
wikingi'll take the ocas stuff16:46
wikingi've assigned myself for those16:46
wikingso if in others the cause is the same then u can skip those16:46
-!- FSCV_ [~FSCV@78.129.244.119] has joined #shogun16:46
-!- blackburn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]16:47
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has joined #shogun16:48
@sonney2kheiko1, how is it going?16:49
heiko1sonney2k: there is a bug in the migration framework16:50
@sonney2kand who did kill blackburn?16:50
heiko1a case I did not think of16:50
heiko1trying to fix for a couple of hours now16:50
@sonney2kheiko1, is it *that* important for this release?16:50
heiko1But I might postpone and remove added parameter16:50
heiko1sonney2k: we cannot add parameters16:50
heiko1this worked only once16:50
wikingheiko1: so no need to run valgrind alone. check the last 5 bugreports on github. you have all the info you need for finding the leak16:50
@sonney2kheiko1, did we do that yet?16:50
heiko1yes16:50
heiko1sonney2k: from version 0 to 116:51
@sonney2kheiko1, ok - but we broke all serialization with 2.0 anyways16:51
@sonney2kand no one complained16:51
heiko1this means we cannot change serialisation parameters without breaking the testsuite16:51
@sonney2kso...16:51
heiko1testuiste ....16:51
heiko1all tests will fail16:51
@sonney2kheiko1, yes but this is for the next release right?16:51
heiko1we could just remove all maps for now ...16:51
heiko1sonney2k: no I added something16:51
@sonney2kwe changed the testsuite anyways.16:52
heiko1we would have to again16:52
@sonney2kheiko1, what?16:52
heiko1sonney2k: the problem is:16:52
heiko1the current testsuite worked16:52
heiko1until I added a variable to multiclassfeatures16:52
heiko1now all examples involving that do not work16:52
heiko1but the migration also does not work16:52
heiko1give me some more minutes, it might be fixed soon16:53
@sonney2kheiko1, 30 mins?16:53
heiko1otherwise I suggest we simply reset, but since I spent so mnuch time on the bug today, I might get rid of it soon16:53
heiko1yes good16:53
wikingwhat is the name of package handler on centos?16:58
@sonney2kwiking, yum17:01
@sonney2kyum install <crap>17:01
wikingcool thnx17:01
wikingmmm17:03
wikingif SGVector is created with: SGVector<float64_t>(samp, m_mean.vlen, false);17:03
wikingwhat is the proper way to call the destructor?17:03
heiko1sonney2k: okay I give up17:03
heiko1sonney2k:  very frustrating to fix17:03
heiko1sonney2k: so now what do you suggest?17:03
heiko1replace whole testsuite?17:03
@sonney2kwiking, ??? you take care of that then17:03
heiko1and remove all parameter maps?17:03
wikingSG_FREE(v.vector)17:03
@sonney2kwiking, yes17:04
wikingok17:04
wikingso either we fix17:04
wikingCGaussian::sample17:04
@sonney2kheiko1, I don't get exactly why you are adding parameters last minute?17:04
@sonney2kwhat happened17:04
@sonney2k?17:04
wikingso it uses ref counter17:04
wikingor we remember this ;)17:04
wikingi would go the latter17:04
@sonney2kwiking, no it doesn't use refcounter17:04
wikingyeah but can't we make it17:04
wikingto use?17:04
@sonney2kthat is why last argument == false17:04
heiko1sonney2k: tried to fix integration tests, Ill undo17:04
@sonney2kheiko1, stop stop17:05
@sonney2kheiko1, I don't get where we are / what the issues are...17:05
heiko1sonney2k:  thing is, I thought the migration worked17:05
heiko1but one important case doesnt17:05
heiko1sonney2k: it is not possible to add parameters without breaking tests17:05
@sonney2kheiko1, what happens then?17:06
heiko1tests fails17:06
heiko1since deserialised data is different17:06
@sonney2kheiko1, well if we add a parameter we have to re-create tests17:06
heiko1adding a parameter map makes the thing just crash17:06
heiko1sonney2k: yes, this is what I wanted to avoid17:06
heiko1thats why we have the migration right?17:06
@sonney2kheiko1, but why? when you add a new parameter (that is to be serialized) then results will change so it is ok!17:07
heiko1sonney2k:   but it doesnt work for one case I forgot when I wrote it17:07
@sonney2kheiko1, the migration was for changing types/renaming variables17:07
@sonney2kbut not *additions*17:07
heiko1and adding17:07
@sonney2kheiko1, but how can an serialized object be the same when you add a parameter?17:08
heiko1sonney2k:  loading doesnt work if the parameter is new17:08
heiko1and the file hasnt got it17:09
@sonney2kheiko1, I see -yes that should work but tests would break anyways even if you fix that17:09
@sonney2kso tests *have to be generated*17:09
-!- SCV__ [~FSCV@118.163.183.203] has joined #shogun17:09
heiko1ok then17:09
heiko1I will remove the "new" mapping for now17:09
@sonney2kheiko1, and in addition we need to fix this (but that is not too important for this release)17:09
heiko1or better17:09
heiko1I will remove serialization17:09
heiko1and then we add this later17:10
@sonney2kwhat?17:10
-!- FSCV_ [~FSCV@78.129.244.119] has quit [Ping timeout: 252 seconds]17:10
@sonney2kwhole serialization?17:10
heiko1of this one parameter17:10
@sonney2kor that yes17:10
heiko1i remove the SG_ADD17:10
@sonney2kyes mark it in a comment17:10
@sonney2kbest solution for now17:10
heiko1sonney2k: i did not expect this to be such a bitch, which is why i started with it17:11
heiko1you know these problems, keep growing the when you invest more time17:11
wikingok17:13
shogun-notifier-shogun: Viktor Gal :master * 807d8ca / src/shogun/features/DataGenerator.cpp: https://github.com/shogun-toolbox/shogun/commit/807d8ca02d81f46f3e87a1216c110e1708b4095c17:13
shogun-notifier-shogun: Fix for issue #929 and #93017:13
wikingdone17:13
shogun-notifier-shogun: Release SGVector gaussian sample in CDataGenerator::generate_gaussians17:13
shogun-notifier-shogun: Viktor Gal :master * b85414b / src/shogun/features/DataGenerator.cpp: https://github.com/shogun-toolbox/shogun/commit/b85414ba0a113b1ac5ced44d4534a8586a23089117:13
shogun-notifier-shogun: Merge pull request #931 from vigsterkr/master17:13
shogun-notifier-shogun:17:13
shogun-notifier-shogun: Fix for issue #929 and #93017:13
-!- blackburn [~blackburn@83.234.169.186] has joined #shogun17:13
wikingtoday was a productive day ;D17:14
wikingblackburn: i think your bug can be closed as well17:14
blackburnwiking: yeah let me check it now17:14
wikinghttps://github.com/shogun-toolbox/shogun/issues/92717:14
wikingsame goes for17:14
wikinghttps://github.com/shogun-toolbox/shogun/issues/92817:14
wikingbut fetch the last commit i've merged and check yourself just in case17:15
shogun-notifier-shogun: Heiko Strathmann :master * 25b9613 / src/shogun/labels/MulticlassLabels.cpp: https://github.com/shogun-toolbox/shogun/commit/25b961322c6acb64a41a47014c722bc6726b927517:15
shogun-notifier-shogun: added parameter mapping for new serialisable parameter which currently maps from 0 to 117:15
shogun-notifier-shogun: Heiko Strathmann :master * fde6304 / src/shogun/machine/KernelMachine.cpp: https://github.com/shogun-toolbox/shogun/commit/fde6304b8ab53733c44f8e13b0c1ede5e86a174717:15
shogun-notifier-shogun: added comments on parameter maps17:15
shogun-notifier-shogun: Heiko Strathmann :master * 9645d2a / examples/undocumented/python_modular/evaluation_clustering.py: https://github.com/shogun-toolbox/shogun/commit/9645d2a8b24252e6920a58ee9c497c8d8328d8bb17:15
shogun-notifier-shogun: removed clustering MI since it does not work with integration tests17:15
shogun-notifier-shogun: Heiko Strathmann :master * d7338cb / src/shogun/labels/MulticlassLabels.cpp: https://github.com/shogun-toolbox/shogun/commit/d7338cb06341b8eef04a1c49e060793cee780c8217:15
shogun-notifier-shogun: removed parameter registration17:15
shogun-notifier-shogun: Heiko Strathmann :master * e651665 / src/shogun/labels/MulticlassLabels.cpp: https://github.com/shogun-toolbox/shogun/commit/e65166525b07863f522c1bd91cb3e5133304a16517:15
shogun-notifier-shogun: forgot this17:15
shogun-notifier-shogun: Heiko Strathmann :master * 9e0747f / / (3 files): https://github.com/shogun-toolbox/shogun/commit/9e0747f4e5fe2d121ae81b1de7422a7cd741340417:15
@sonney2kheiko1, yeah that is why we really need such hard feature freezes17:15
shogun-notifier-shogun: Merge pull request #932 from karlnapf/master17:15
shogun-notifier-shogun:17:15
shogun-notifier-shogun: removed parameter registration for multiclass labels again17:15
blackburnoh my f17:15
blackburng17:16
heiko1sonney2k: blackburn, wiking I need a break, will be back in a little while17:16
blackburnheiko1 goes crazy17:16
@sonney2kblackburn, heiko1 what is the status then?17:16
@sonney2kwhat is missing?17:16
heiko1loads of bugs :)17:16
blackburnsonney2k: a few fixes for unit-tests I think17:16
blackburnhah17:16
@sonney2khaha17:16
blackburnthat's always the case nevermind17:16
heiko1integration also dont work all17:16
heiko1but more that before17:16
blackburnheiko1: the evolution17:16
heiko1but I think its fine to release once the tests work again17:16
@sonney2kheiko1, blackburn my suggestion would be17:17
@sonney2kto disable all (crashing/not working) integration tests17:17
blackburnlets follow darwin stuff17:17
blackburnand hope we will get better with time17:17
blackburn:D17:17
@sonney2kthen on the buildbot create build failues when these tests fail17:17
blackburnsonney2k: yes disable all converters in particular17:17
heiko1agreed17:17
heiko1getting a coffee.....17:17
@sonney2kand then gradually try to improve the situation for later17:17
-!- SCV__ [~FSCV@118.163.183.203] has quit [Ping timeout: 264 seconds]17:20
shogun-buildbotbuild #781 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/781  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Viktor Gal <viktor.gal@maeth.com>17:23
-!- SCV__ [~FSCV@204.45.132.11] has joined #shogun17:26
wikingblackburn: bugs are closed... they are fixed17:27
blackburnwiking: alright thanks17:27
wikingnw17:27
shogun-buildbotbuild #782 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/782  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Viktor Gal <vigsterkr@gmail.com>17:27
blackburnone bug to rule them all17:27
wikingindeed17:31
shogun-buildbotbuild #783 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/783  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>17:31
wikingso now only customkernel and combinedkernel...17:31
* sonney2k wonders what you guys are doing17:33
blackburnsonney2k: valgrinding unit-tests17:33
wikingyeah that's the ultimate combo17:34
wiking:D17:34
@sonney2kwiking, wait so it is default now?17:34
@sonney2kvalgrinded unit tests?17:34
wikingsonney2k: nono17:35
wikingmake valgrind-per-module17:36
wikingthat generates the per unit test module valgrind output17:36
wikingof course under tests/unit17:36
@sonney2kwiking, blackburn I let you guys fix the tests...17:37
blackburnsonney2k: thanks! :)17:38
wikingimo this fails http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/783/steps/test/logs/stdio17:38
wikingbecause of a missing macro test17:38
blackburnwiking: eigen3?17:39
wikingdunno17:40
@sonney2kblackburn, should the SPE with local strategy window be empty in converter_spe_helix.py ?17:42
blackburndoes it have a window?17:43
blackburnitneresting17:43
blackburnsonney2k: it is not empty here hmm17:43
@sonney2kblackburn, 3/4 have some content17:44
blackburnsonney2k: that's a bug for sure but why..17:45
@sonney2kinteractive_svr_demo.py doesn't work17:45
@sonney2kso_multiclass_BMRM.py neither17:49
@sonney2kthe rest is good!17:50
blackburnsonney2k: yeah bmrm is quite strange though17:52
heiko1back17:58
heiko1sonney2k: blackburn, I would even suggest to freeze features even after the release to get everything in shape for gsoc17:59
@sonney2kheiko1, nak - we should rather change our development model and no longer work on master18:06
@sonney2kbut on dev branches then18:06
heiko1sonney2k: explain18:06
@sonney2kand once sth is stable enough merge the dev branch over to master18:07
heiko1sonney2k: I am just thinking of a way of getting everything to work at some point18:07
@sonney2kheiko1, haha :)18:07
heiko1sonney2k: at least the known issues :)18:07
heiko1sonney2k: like this idea, have the master branch always clean18:07
heiko1dev branch is used for new stuff18:07
heiko1then this stuff is being polished, and once it works we merge it18:08
heiko1this means that we currently only have a "dev" branch and no master branch :)18:08
@sonney2kheiko1, IIRC last year someone described in very detail how one should do it18:14
@sonney2kwe just didn't follow the process18:14
heiko1who?18:15
@sonney2kforgot18:15
@sonney2ksomeone onthe mailinglist18:15
heiko1I guess its also a matter of discipline18:15
heiko1no one likes to fix these horrible bugs in our system18:15
heiko1its cooler to write new ML algos18:15
shogun-buildbotbuild #920 of deb3 - modular_interfaces is complete: Failure [failed test python_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/920  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Viktor Gal <vigsterkr@gmail.com>18:16
heiko1sonney2k:  so whats currently going on?18:17
@sonney2kheiko1, no idea...18:17
heiko1fixing tests and then go?18:17
heiko1I mean unit/examples18:17
@sonney2kheiko1, I am fixing 1 graphical example currently18:17
heiko1sonney2k: I see18:18
@sonney2kbut the others - blackburn, wiking?18:18
heiko1I guess I could try to fix some more integration tests then18:18
heiko1the obvious ones18:18
heiko1I know there are a few18:18
@sonney2kheiko1, feel free18:18
-!- FSCV_ [~FSCV@vpnreactor-static44.den.icastcenter.com] has joined #shogun18:19
-!- SCV__ [~FSCV@204.45.132.11] has quit [Ping timeout: 260 seconds]18:20
shogun-notifier-shogun: Viktor Gal :master * 286c078 / tests/unit/regression/GaussianProcessRegression_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/286c078b8147b22e3155b0092c06bc2b3604937518:21
shogun-notifier-shogun: Add HAVE_EIGEN3 check to GaussianProcessRegression_unittest.cc18:21
shogun-notifier-shogun: Viktor Gal :master * 333fce5 / tests/unit/regression/GaussianProcessRegression_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/333fce5b7e7472d396f4146a7f1c2a2d7d8fd72018:21
shogun-notifier-shogun: Merge pull request #933 from vigsterkr/master18:21
shogun-notifier-shogun:18:21
shogun-notifier-shogun: Add HAVE_EIGEN3 check to GaussianProcessRegression_unittest.cc18:21
shogun-notifier-shogun: Heiko Strathmann :master * 4da2a23 / examples/undocumented/libshogun/ (2 files): https://github.com/shogun-toolbox/shogun/commit/4da2a23b474958427855af7a978f332d3649cc7818:21
shogun-notifier-shogun: added an example for multiclass serialization that doesnt work (but runs fine) Once muticlass label serialization work this should be turned into a test18:21
shogun-notifier-shogun: Heiko Strathmann :master * b51ee30 / examples/undocumented/libshogun/serialization_multiclass_labels.cpp: https://github.com/shogun-toolbox/shogun/commit/b51ee305b33fd53544f4ed02e3e38b906f1d21b818:21
shogun-notifier-shogun: added a test for multiclass serialization18:21
shogun-notifier-shogun: Heiko Strathmann :master * 0aa25e8 / examples/undocumented/libshogun/labels_multiclass_serialization.cpp: https://github.com/shogun-toolbox/shogun/commit/0aa25e8b24d303519c7692a95ed3d6e68ca4f7c318:21
shogun-notifier-shogun: removed wrongly named file18:21
shogun-notifier-shogun: Heiko Strathmann :master * b2313d0 / tests/unit/base/Serialization_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/b2313d0d1b38b94686621e5a749b9a1f3c95674018:21
shogun-notifier-shogun: added a provisory test for multiclass label serialization which has to be activated once this works18:21
shogun-notifier-shogun: Heiko Strathmann :master * 4110afc / examples/undocumented/libshogun/ (2 files): https://github.com/shogun-toolbox/shogun/commit/4110afcb7cda05f29564691076ae20cdd4c172f318:21
shogun-notifier-shogun: added an example for adding new parameters to classes. However this one does not work, must be activated once it works18:21
shogun-notifier-shogun: Heiko Strathmann :master * 1459ba0 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/1459ba0a45afd4700a43709d91986b94f62ec48918:21
shogun-notifier-shogun: Merge pull request #934 from karlnapf/master18:21
shogun-notifier-shogun:18:21
shogun-notifier-shogun: added a collection of tests/exam?ples for things that currently do not work18:21
@sonney2kheiko1, examüples?18:21
@sonney2k*lol*18:21
heiko1sonney2k: its what I did today18:22
heiko1not activated though18:22
heiko1just to illustrate the bug18:22
-!- blackburn [~blackburn@83.234.169.186] has left #shogun []18:22
* sonney2k loves examüples18:22
heiko1I reference it in the issue I am currently writing18:22
wikingsonney2k: here?18:24
heiko1wiking:  trying your new make18:26
wikingheiko1: should work18:26
wikingsonney2k: 2 things: could you please change on ubu1 from -j4 to -j2 please?18:26
heiko1wiking: it works!18:26
heiko1nice18:26
wikingand 2) could we switch from gcc to clang on bsd1?18:26
heiko1reveals many memory errors :|18:27
wikingheiko1: still?18:27
heiko1wiking: yeah, but thats what its for right?18:27
wikingyeah18:27
wikingbut i dont get now why this shit fails on bsd still18:27
shogun-buildbotbuild #784 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/784  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>18:27
wikinghttp://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/784/steps/test/logs/stdio18:27
heiko1wiking:  that error is weird, I dont have it here, is it because of eigen?18:28
wikingyes18:28
wikingbut imo i fixed with my last commit18:29
heiko1oh18:29
heiko1maybe a pending build?18:29
-!- FSCV_ [~FSCV@vpnreactor-static44.den.icastcenter.com] has quit [Ping timeout: 240 seconds]18:29
wikingletsee18:29
wikingyea maybe the new 785 build18:29
wikingyeah that includes my patch as well18:30
shogun-buildbotbuild #785 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/785  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Viktor Gal <vigsterkr@gmail.com>18:31
wikingdoh18:32
wikingafaik same problem ExactInferenceMethod_unittest.cc18:32
wiking:)18:32
shogun-notifier-shogun: Soeren Sonnenburg :master * 5cd042d / / (2 files): https://github.com/shogun-toolbox/shogun/commit/5cd042d83495f5e1789d2ff387809326b378332018:33
shogun-notifier-shogun: fix ocr & svr demo18:33
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun18:35
@sonney2kn4nd0, hey there!18:35
n4nd0sonney2k: hello hello18:35
@sonney2kn4nd0, and now the attack!18:36
n4nd0:)18:36
-!- blackburn [~blackburn@83.234.169.186] has joined #shogun18:36
shogun-notifier-shogun: Viktor Gal :master * d06ebf4 / tests/unit/regression/gp/ExactInferenceMethod_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/d06ebf49996c36147047060c754d62a0d1688df518:36
shogun-notifier-shogun: Add check for HAVE_EIGEN3 in ExactInferenceMethod_unittest18:36
shogun-notifier-shogun: Viktor Gal :master * 7f80fe7 / tests/unit/regression/gp/ExactInferenceMethod_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/7f80fe76d5c9b0f97d90fa3c3c6708abcada5f6618:36
shogun-notifier-shogun: Merge pull request #936 from vigsterkr/master18:36
shogun-notifier-shogun:18:36
shogun-notifier-shogun: Add check for HAVE_EIGEN3 in ExactInferenceMethod_unittest18:36
n4nd0sonney2k: tell me18:36
@sonney2kdo you know why so_multiclass_BMRM.py complains about18:36
@sonney2kbmrm_return_value_T?18:36
shogun-buildbotbuild #786 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/786  blamelist: Viktor Gal <viktor.gal@maeth.com>18:36
n4nd0sonney2k: mmm no, no idea what that parameter is18:36
@sonney2kn4nd0, I've recently changed the .i files to not include the bmrm_return_value_T18:36
@sonney2kso I am wondering if we return this somehow in python interfaces18:36
blackburnsonney2k: we use it in interfaces unfortunately18:37
n4nd0sonney2k: ok, why did you change it not to include it?18:37
@sonney2kn4nd0, btw sitting in c-base currently...18:37
blackburnlet me get it back18:37
@sonney2kblackburn, where when why?18:37
@sonney2kblackburn, not yet please18:37
n4nd0sonney2k: aham! are you going to talk to the people there??18:37
blackburnsonney2k: so_multiclass_BMRM18:37
blackburnand so_multiclass_BMRM_director18:37
n4nd0about the workshop18:37
@sonney2kn4nd0, I hope I can today18:37
@sonney2kn4nd0, yes18:37
n4nd0nice18:37
@sonney2kif I don't fall asleep before it starts...18:38
blackburnsonney2k: does that happen you fall asleep in public?18:38
blackburn:D18:38
@sonney2kblackburn, today yes!18:39
@sonney2kI am dead tired!18:39
@sonney2kblackburn, where exaclty is that bmrm_return_value_T used?18:39
blackburnsonney2k: graphical/so_multiclass_director_BMRM.py:12118:39
@sonney2kblackburn, in here I mean so_multiclass_BMRM.py?18:40
shogun-notifier-shogun: Heiko Strathmann :master * 5b7f8d7 / tests/unit/regression/GaussianProcessRegression_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/5b7f8d7e7ae0ab703762714d23aefe0e43354d1218:40
shogun-notifier-shogun: Squashed commit of the following:18:40
shogun-notifier-shogun:18:40
shogun-notifier-shogun: commit 20fca14ad5f4bfd1f618cb9dd58287828255c21918:40
shogun-notifier-shogun: Author: Heiko Strathmann <heiko.strathmann@gmail.com>18:40
shogun-notifier-shogun: Date:   Thu Mar 14 17:38:03 2013 +000018:40
shogun-notifier-shogun: Heiko Strathmann :master * a33c97c / tests/unit/regression/GaussianProcessRegression_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/a33c97c4334dc7ca0085fb1e67a722d216d5b3c018:40
shogun-notifier-shogun: Merge pull request #937 from karlnapf/master18:40
shogun-notifier-shogun:18:40
n4nd0blackburn: does it fail on tests?18:40
shogun-notifier-shogun: fixed and enabled GPR unit test18:40
blackburnsonney2k: 7318:40
blackburnn4nd0: what?18:40
n4nd0blackburn: the complaint sonney2k talked about, where does it happen?18:41
n4nd0about bmrm_return_value_T18:41
blackburnn4nd0: so_multiclass_BMRM.py:7218:41
heiko1sonney2k:  hey cool the interactive gp demo now works :)18:41
blackburnsonney2k: I can't see where it is ignored18:42
n4nd0\o/18:42
@sonney2kI see get_result()18:42
blackburnsonney2k: it is of bmrm_return_value_t18:43
@sonney2kyeah then we need it18:43
@sonney2kblackburn, could we rather move the file to say shogun/structure/libbmrm_result.h?18:44
blackburnsonney2k: I do not mind, let me do that18:44
@sonney2kblackburn, could we rather move the file to say shogun/structure/libbmrm_return_value.h?18:44
@sonney2kI meant18:44
@sonney2kand then include that in the libbmrm.h18:45
@sonney2k(which we then don't have to include in the .i)18:45
@sonney2kblackburn, thanks18:45
@sonney2kblackburn, when again is the gsoc application deadline18:46
@sonney2k?18:46
blackburnsonney2k: 2918:46
@sonney2kblackburn, so we need to have it ready *next* week18:46
blackburnsonney2k: yes18:46
blackburnthat's ok for me18:46
blackburnsonney2k: looks quite easy - I'll just replace all these kill/not-kill with something more reasonable18:47
n4nd0haha18:48
blackburnn4nd0: tapkee is dead under shindows18:50
blackburndue to ezoptionsparser18:50
n4nd0blackburn: its dependencies or?18:51
blackburnn4nd0: that's how we parse options :)18:51
n4nd0I meant ezoptionsparser dependencies18:51
n4nd0not tapkee's18:51
blackburnn4nd0: ah no idea18:51
blackburnits code18:52
n4nd0aham18:52
n4nd0blackburn: should we try to fix it or skip it?18:52
blackburnn4nd0: later18:53
shogun-notifier-shogun: Sergey Lisitsyn :master * 91d957d / src/shogun/lib/tapkee/routines/multidimensional_scaling.hpp: https://github.com/shogun-toolbox/shogun/commit/91d957d412f4f323bbb84f14243540d0494849b118:54
shogun-notifier-shogun: Hopefully last update of tapkee before release18:54
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 246 seconds]18:56
@sonney2kn4nd0, about the website menu - I run you test site18:57
@sonney2kdo I have to copy the extra files into that?18:57
n4nd0sonney2k: no18:57
@sonney2kn4nd0, I don't see a menu yet I mean18:57
@sonney2kno drop down or anything18:57
n4nd0sonney2k: the extra files where the integration into shogun18:57
n4nd0sonney2k: what page are you opening?18:58
@sonney2kn4nd0, err python manage.py runserver18:58
n4nd0sonney2k: do you get an error you mean?18:59
@sonney2kn4nd0, no no error but just the site you showed me a few weeks ago18:59
n4nd0sonney2k: open drop18:59
@sonney2kbut no menu - or didn't you start that yet?18:59
n4nd0http://127.0.0.1:8000/drop18:59
wikingblackburn: why tapkee is not as a module in shogun?19:00
blackburnwiking: module like?19:00
wikingblackburn: git submodule19:00
n4nd0sonney2k: see it now? it is a bit buggy though19:00
blackburnwiking: that's quite difficult as there are other stuff like build system, etc19:01
wikingblackburn: ic19:01
@sonney2kn4nd0, I see it19:01
@sonney2kdid you try to get this into the website yet?19:01
n4nd0sonney2k: yes19:01
@sonney2kn4nd0, where?19:01
n4nd0sonney2k: the other two files I attached in the mail19:01
n4nd0sonney2k: are my guess19:02
n4nd0sonney2k: but it doesn't work fine19:02
@sonney2kn4nd0, I am a bit lost - what do I have to copy where to start?19:02
n4nd0the css of the current webpage and of if dropdown menu have to be merged somehow19:02
@sonney2kn4nd0, yeah I want to do that but I need to know what I have to copy where to get it to start - otherwise I have to start from scratch19:03
n4nd0in addition, I think that it is an issue for the dropdown menu that the current menu we have (the thick black bar in here http://shogun-toolbox.org/page/about/information for example) is an image19:03
n4nd0sonney2k: give a moment to tell you where to copy19:04
n4nd0sonney2k: from the files I attached in the mail, page.html goes under templates and subpage.html too19:05
n4nd0and the javascripts under static/js/19:05
n4nd0sonney2k: I did some experiments trying to merge the css files but I didn't attach those because things were getting more screwed up19:06
n4nd0sonney2k: the dropdown menu's css is in mysite/drop/static/drop.css19:07
shogun-buildbotbuild #921 of deb3 - modular_interfaces is complete: Failure [failed test ruby_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/921  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>19:14
shogun-notifier-shogun: Heiko Strathmann :master * 35cbc51 / tests/unit/regression/GaussianProcessRegression_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/35cbc5149b6e997ba5c38c1250c652c0ab4b851b19:14
shogun-notifier-shogun: fixed memory errors and added asserts again19:14
shogun-notifier-shogun: Heiko Strathmann :master * 20fca14 / tests/unit/regression/GaussianProcessRegression_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/20fca14ad5f4bfd1f618cb9dd58287828255c21919:14
shogun-notifier-shogun: really fixed it now19:14
shogun-notifier-shogun: Heiko Strathmann :master * 4aad854 / examples/undocumented/python_modular/ (2 files): https://github.com/shogun-toolbox/shogun/commit/4aad85416f02a54391ae57a3d7fa62a6847ce0ae19:14
shogun-notifier-shogun: made two more integration tests work19:14
shogun-notifier-shogun: Heiko Strathmann :master * 2436f2e / examples/undocumented/python_modular/regression_kernel_ridge_modular.py: https://github.com/shogun-toolbox/shogun/commit/2436f2e684be12153d3602bd388f1a2824aeddf819:14
shogun-notifier-shogun: updated example to work with data that makes sense19:14
shogun-notifier-shogun: Heiko Strathmann :master * f5d4943 / examples/undocumented/python_modular/ (3 files): https://github.com/shogun-toolbox/shogun/commit/f5d494345bbbbec4aa85107b912a3abf70f75e1f19:14
shogun-notifier-shogun: Merge branch 'bugsprint'19:14
shogun-notifier-shogun: Sergey Lisitsyn :master * a697aab / src/shogun/lib/tapkee/routines/multidimensional_scaling.hpp: https://github.com/shogun-toolbox/shogun/commit/a697aabc67519c05cc478c0202bbd52e618b5a6419:14
shogun-notifier-shogun: Hopefully last update of tapkee before release19:14
shogun-notifier-shogun: Heiko Strathmann :master * 64d20d4 / examples/undocumented/python_modular/ (3 files): https://github.com/shogun-toolbox/shogun/commit/64d20d4bb79a26c832872d5d32760707de89915a19:14
shogun-notifier-shogun: Merge pull request #938 from karlnapf/master19:14
shogun-notifier-shogun:19:14
shogun-notifier-shogun: more integration test fixes19:14
heiko1sonney2k: why are all regression tests done with non-regression data?19:15
n4nd0I got internal c++ compiler error building shogun :D19:15
@sonney2k n4nd0 haha19:15
n4nd0and now virtual memory exhausted: Cannot allocate memory19:16
n4nd0will try without ccache..19:16
@sonney2kheiko1, our examples are not reasonable in many cases...19:16
heiko1n4nd0: uh ....19:16
blackburnsonney2k: still getting no destructor found19:16
@sonney2kn4nd0, how much memory do you have?19:16
heiko1sonney2k: aksing why, its so much easier to check whether a method is working when simple but meaningful toy data is used19:16
blackburnmay be virtual19:16
n4nd0sonney2k: 4GB RAM, and 1GB swap which is never used19:16
n4nd0but my RAM monitor tells me everything was not used19:17
@sonney2kheiko1, would be cool to have!19:17
n4nd0I have no idea why that is happening19:17
n4nd0will clean and build again19:17
@sonney2killustrative examples that could even generate nice pictures - a long standing dream19:17
heiko1sonney2k: yeah, at least we should do this from now :)19:17
@sonney2kn4nd0, well shogun will need 3-4GB  :)19:18
heiko1whenever i replace an integration test I have to make sure the things works anyway, and then I just replace data by something useful19:18
@sonney2kheiko1, do it!19:18
n4nd0sonney2k: mmm I have built it here several times now so I guess something could be wrong because I didn't clean after rebasing19:20
shogun-buildbotbuild #787 of bsd1 - libshogun is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/78719:23
@sonney2kblackburn, did you include that file in the .i before the other SO methods?19:24
blackburnsonney2k: yeah that can be the reason checking19:24
blackburnsonney2k: so are we releasing today?19:25
heiko1blackburn:  cou you check converter_multidimensionalscaling_modular.py and classifier_multiclassocas_modular.py for integration tests?19:26
heiko1and replace tests?19:26
blackburnheiko1: alright19:26
shogun-buildbotbuild #922 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/92219:28
shogun-notifier-shogun: Heiko Strathmann :master * 36120de / examples/undocumented/python_modular/ (3 files): https://github.com/shogun-toolbox/shogun/commit/36120ded54602981eb32a53174d09589d3c15ae319:31
shogun-notifier-shogun: more working integration tests19:31
shogun-notifier-shogun: Heiko Strathmann :master * 5a9e1c8 / examples/undocumented/python_modular/ (3 files): https://github.com/shogun-toolbox/shogun/commit/5a9e1c817f88e9f7266888f84d5634436741bf3019:31
shogun-notifier-shogun: Merge pull request #939 from karlnapf/master19:31
shogun-notifier-shogun:19:31
shogun-notifier-shogun: more integration test fixes19:31
heiko1sonney2k: blackburn, itegration test errors slowly going to zero19:32
heiko1blackburn:  and preprocessor_dimensionreductionpreprocessor_modular.py19:35
blackburnok19:36
heiko1sonney2k:  blackburn, whats the state?19:46
blackburnheiko1: fighting with bmrm return value19:46
blackburnah works19:47
blackburnice19:47
blackburnheiko1: will fix integration stuff very soon19:48
blackburnjust have to finish so examples19:48
-!- FSCV [~FSCV@76.73.3.220] has joined #shogun19:57
heiko1sonney2k: blackburn I am very tired (all day hackathon), going home now and will come in later to check how its going20:02
blackburnheiko1: sure, take a rest20:02
-!- heiko1 [~heiko@nat-167-251.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]20:08
shogun-notifier-shogun: Sergey Lisitsyn :master * 436952b / / (6 files): https://github.com/shogun-toolbox/shogun/commit/436952b23991277a31418a1668195c3310e5b67420:15
shogun-notifier-shogun: Fixed bmrm_return_value issues20:15
shogun-notifier-shogun: Sergey Lisitsyn :master * b9c4d1a / data: https://github.com/shogun-toolbox/shogun/commit/b9c4d1a227a8f511839e08b58ec12fdf471093f520:15
shogun-notifier-shogun: updated data20:15
shogun-notifier-shogun: Sergey Lisitsyn :master * e8ecefe / data: https://github.com/shogun-toolbox/shogun/commit/e8ecefe4f27a4716800414e64f374070e45bb30c20:32
shogun-notifier-shogun: Updated data20:32
shogun-buildbotbuild #925 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/925  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>20:42
@sonney2kblackburn, did you introduce getters for get_hist_Fd etc?20:51
@sonney2kblackburn, there is the issue that we need the variables to be private/protected then.20:52
blackburnsonney2k: doing20:52
@sonney2kblackburn, I guess you will need setters then too...20:53
blackburnsonney2k: yeah20:58
shogun-buildbotbuild #926 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/926  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>20:59
@sonney2kblackburn, are you working on it?21:02
blackburnsonney2k: yes21:02
blackburnsonney2k: quite a lot changes21:03
@sonney2kyeah I would expect so :/21:04
blackburnsonney2k: did you talk to c-base?21:05
@sonney2kblackburn, sitting in the meeting21:05
blackburnoh I see21:05
@sonney2kwill be next in the slot...21:05
blackburnsonney2k: sounds like a bureaucracy21:06
blackburnsonney2k: that's something wrong with what I do now21:09
blackburnsonney2k: I'd rather rename fields that are assumed to be private21:10
blackburnor rather rename getters21:11
wikinghahahahahahahhaha21:14
wikingahahahahah21:14
wikingblackburn: http://i.imgur.com/haspR.jpg21:14
blackburnwiking: not possible I guess :)21:15
wikingblackburn: well it seems it is21:15
wiking:)21:15
wikingzwolnij means release :DD21:17
blackburnwiking: heh21:17
@sonney2keverything positive here from the c-base side!21:19
wiking\o/21:20
blackburnnice21:20
shogun-buildbotbuild #927 of deb3 - modular_interfaces is complete: Failure [failed compile csharp_modular]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/927  blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com>21:21
shogun-notifier-shogun: Sergey Lisitsyn :master * a03730b / / (3 files): https://github.com/shogun-toolbox/shogun/commit/a03730bd045359e64be67436927da228501877e621:22
shogun-notifier-shogun: Resolved a conflict with BMRM stuff21:22
-!- sumit [ca4eaca2@gateway/web/freenode/ip.202.78.172.162] has quit [Ping timeout: 245 seconds]21:48
-!- heiko [~heiko@027d8b62.bb.sky.com] has joined #shogun22:10
heikosonney2k: blackburn whats up?22:12
blackburnheiko: I fixed bmrm stuff22:15
blackburnheiko: ah and c-base is positive22:15
heikopositive?22:15
blackburnheiko: not in ml mean22:16
blackburnheiko: positive about our drinkup22:16
blackburn:D22:16
heikohaha :)22:16
heikook22:16
heikoblackburn: so I dont know what we need to do to release22:16
heikowhats missing etc22:16
heikodo you?22:16
blackburnheiko: I don't either22:16
blackburnif csharp compiles lets release tomorrow22:17
heikoI am up for it22:17
blackburnas sonney2k is 'dead'22:17
heikookay22:17
heikoI see22:17
heikowell lets just wait for him22:17
heikoblackburn: do you have any ideas how to get rid of those loads of bugs?22:17
blackburnheiko: rewrite it22:17
blackburnrewrite everything22:17
heiko?22:18
heikoserious?22:18
blackburnheiko: jk22:18
blackburnheiko: tests is the only way I guess22:18
heikowe need to somehow make ourselves attach this22:18
heikoblackburn: what do you think on a more extended feature freeze time?22:18
heikowhere all efforts go into fixing things22:18
blackburnheiko: I'd like to start with feature drop time22:18
heikonamely?22:18
blackburnheiko: that's to be decided22:19
blackburnheiko: the codebase is huge - untestable22:19
heikoblackburn: I agree22:19
heikomaybe we could have some kind of automated extration of the stable stuff22:19
blackburnheiko: we could move some code out of shogun as shogun extensions22:20
heikolike: only if there exist examples/unit-tests/integration tests for things, then it is added to stable22:20
shogun-buildbotbuild #928 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/92822:20
blackburncsharp compiled, nice22:20
heikonice22:21
heikoblackburn:  but what happens to it then?22:22
blackburnheiko: no idea22:22
heikoblackburn: I feel the only way to make things stable is to make sure "they will never break again"22:22
heikoyou know, then one could relax about things not working22:22
heikoand if we then also did this advanced dev-process that sonney2k mentioned, we could detect everything22:22
heikobut if I say a certain part is "finished", it should work forever22:23
heikofor my feeling we do way to much fixing which lasts for a short time22:23
heikoblackburn:  and yes only way for this is having tests, but I guess it would also change the development process22:24
heikolike you said, first tests, then code22:24
heikoto have a 100% coverage etc22:24
blackburnheiko: it depends what we aim..22:24
heikoblackburn: true22:24
blackburnheiko: we have no 'vector'22:24
heiko?22:24
heikoexplain22:25
blackburnheiko: I do not know where we are moving to22:26
heikowell maybe we should think of this22:26
blackburnheiko: we should for sure - and discuss it at our meeting may be22:27
heikoblackburn: I realize more and more that I want to do stable software, all this bugfixing annoys me too much22:27
heikoand I am including my own code in this22:27
heikoits just, I dont know how to change this without starting from scratch22:28
heikotake the migration framework22:28
heikoI spent like a few hundred hours on this22:28
heikoand now?22:28
heikoit doesnt work22:28
heikoan I have to think myself into it again22:28
heikoor the GPR22:28
heikosome guy spent a whole summer on this, and then it just doesnt work22:29
blackburnheiko: yes that's quite bad22:29
heikowhat a waste of time22:29
heikoBut I feel like the only way to get rid of this is to rigorously change the way things are developed22:29
heikowhich is slow and time consuming22:29
heikoon the other hand one saves so much time since things do not just silently break22:30
heikoblackburn: anyway, I leave you alone now :D22:30
heikoI will check back tomorrow see what the release says22:30
blackburnheiko: back22:44
blackburnheiko: yeah that's a mistake - you write only code and save time22:45
blackburnbut in the end you spend time fixing bugs22:45
blackburn(or the code doesn't work at all)22:45
blackburnoh my god23:43
blackburnn4nd0: I was just contacted by guy who is going to publish a paper mentioning tapkee23:43
--- Log closed Fri Mar 15 00:00:12 2013
« Wednesday, 2013-03-13 Index Friday, 2013-03-15 »

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!