IRC logs of #shogun for Saturday, 2015-04-04

« Friday, 2015-04-03 Index Sunday, 2015-04-05 »
--- Log opened Sat Apr 04 00:00:28 2015
-!- kshitijk [~kshitijk@220.227.149.86] has quit [Ping timeout: 255 seconds]00:10
-!- AlexLuya [~alex@2403:d400:1000:1:9c0:327f:39e7:fd1c] has joined #shogun04:16
-!- AlexLuya [~alex@2403:d400:1000:1:9c0:327f:39e7:fd1c] has quit [Quit: Leaving]04:46
-!- pickle27 [~pickle27@192-0-134-235.cpe.teksavvy.com] has joined #shogun05:27
-!- pickle27 [~pickle27@192-0-134-235.cpe.teksavvy.com] has quit [Remote host closed the connection]06:00
-!- kshitijk [~kshitijk@220.227.149.86] has joined #shogun06:33
-!- kshitijk [~kshitijk@220.227.149.86] has quit [Ping timeout: 272 seconds]06:58
-!- besser82 [~besser82@fedora/besser82] has joined #shogun08:32
-!- mode/#shogun [+o besser82] by ChanServ08:32
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun10:04
shogun-notifier-shogun: Bj?rn Esser :bugfix/python_modular_swig3 * 2067632 / examples/undocumented/python_modular/ (2 files): https://github.com/shogun-toolbox/shogun/commit/2067632151fee2c60d239a2c4dbc4fdcfb34ae6d10:04
shogun-notifier-shogun: fix failing test when building with swig 3.X10:04
shogun-notifier-shogun: Bj?rn Esser :bugfix/python_modular_swig3 * 6e632cb / src/shogun/lib/memory.cpp: https://github.com/shogun-toolbox/shogun/commit/6e632cb3228102f516210bcbcacfde9bd8e821d910:04
shogun-notifier-shogun: fix segfault when building with 3.X in 2 python_modular-tests10:04
shogun-notifier-shogun:   * python_modular-evaluation_cross_validation_multiclass_storage10:04
shogun-notifier-shogun:   * python_modular-evaluation_multiclassovrevaluation_modular10:04
-!- kshitijk [~kshitijk@220.227.149.86] has joined #shogun10:29
-!- besser82 [~besser82@fedora/besser82] has quit [Remote host closed the connection]11:28
-!- nszceta [~nszceta@5ED05E57.cm-7-1b.dynamic.ziggo.nl] has joined #shogun11:48
-!- besser82 [~besser82@fedora/besser82] has joined #shogun12:09
-!- mode/#shogun [+o besser82] by ChanServ12:10
@besser82lisitsyn, g'morning!  :D12:10
@lisitsynbesser82: hey12:10
@besser82lisitsyn, you can read gdb output? :P12:11
@lisitsynlldb yes12:11
@besser82lisitsyn, that basically the same I think ;)12:12
@lisitsynnot really the same but in some essential parts yeah12:12
@besser82lisitsyn, I just have some disturbing segfault left with swig312:12
@lisitsynso not all segfaults are gone yet?12:13
@besser82lisitsyn, all but just one12:13
@lisitsynI see12:14
@besser82lisitsyn, 2 tests are segfaulting...  both related to evaluation-multiclass-bullshit  :P12:14
@besser82lisitsyn, http://fpaste.org/207167/12:15
@besser82lisitsyn, and both are caused by shogun::sg_generic_free<shogun::SGMatrix<double>12:15
@lisitsynha12:16
@besser82lisitsyn, vhat? ^^12:16
@lisitsynthis could be not swig12:16
@besser82lisitsyn, only with swig 3.x12:17
@lisitsynah12:17
@besser82lisitsyn, on swig 2.0.12 all is fine over there12:17
@besser82lisitsyn, so I'm basically a bit puzzled about the why...12:18
@besser82lisitsyn, and where and however12:18
@besser82lisitsyn, that's the last shiatz with da swig312:18
@besser82lisitsyn, everything else ist pretty12:18
@besser82lisitsyn, with just doing that stupid one-liner-patch in swig :P12:19
@lisitsynbesser82: I am glancing over the code12:19
@besser82lisitsyn, okie :D12:20
@besser82lisitsyn, I didn't see anything obvious though12:20
@besser82lisitsyn, hit me, when rdy ;)12:20
@lisitsynbesser82: these inplace ctors look pretty suspicious12:21
@besser82lisitsyn, where exactly?12:21
@lisitsynin CCrossValidationMulticlassStorage12:21
@besser82lisitsyn, cpp?  which line number?12:22
@lisitsynbesser82: see dtor at 3412:22
@lisitsynand init at 6412:22
@lisitsynthey look like dangerous12:22
@besser82lisitsyn, like the SG_FREE parts?12:25
@besser82lisitsyn, or the SG_UNREF?12:26
@lisitsynbesser82: I am more concerned about12:26
@lisitsynm_conf_matrices[i].~SGMatrix<int32_t>();12:26
@lisitsynand new (&m_fold_ROC_graphs[i]) SGMatrix<float64_t>();12:26
@besser82lisitsyn, how to fix that shit?12:27
@besser82lisitsyn, that one fixed and swig3-stuff is done :D12:30
@lisitsynbesser82: don't know :)12:33
-!- nszceta [~nszceta@5ED05E57.cm-7-1b.dynamic.ziggo.nl] has quit [Quit: My Mac has gone to sleep. ZZZzzz...]12:41
@besser82lisitsyn, no ideas?  :(12:52
@besser82lisitsyn, I'll recompile with -O0 and check output again :P12:56
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]13:04
@besser82lisitsyn, trying again with src/shogun/evaluation/CrossValidationMulticlassStorage.cpp:41 commented out...  let's see the outcome...13:31
@lisitsynbesser82: ok I am back sorry13:35
@lisitsynbesser82: should work13:35
@lisitsynthe question is why13:35
@besser82lisitsyn, think so too...13:35
@besser82lisitsyn, there is nothing real problematic found be me so far13:36
@besser82~SGMatrix<T> just calls SGReferencedData::unref()13:36
@lisitsynbesser82: either it wasn't initialized13:37
@lisitsynor it is freed twice13:37
@besser82lisitsyn, or?13:37
@besser82lisitsyn, freeing twice shouldn't be the problem13:37
-!- PirosB3 [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has joined #shogun13:37
@lisitsynbesser82: I see a flaw here anyway13:37
@lisitsyndtor is called always13:38
@lisitsynbut this thing is initialized *sometimes*13:38
@besser82lisitsyn, and not always?13:38
@lisitsynyeah13:38
@besser82lisitsyn, mhhh...  that might be the problem then...13:38
@lisitsynbesser82: line 3613:39
@besser82lisitsyn, so we need some fscking conditional around or make sure it is always initialized13:39
@lisitsynif (compute_ROC && m_fold_ROC_graphs)13:39
@lisitsyncould you try that?13:39
@lisitsynand lines 44, 5213:39
@besser82lisitsyn, just a sec... ;)13:39
@lisitsynI was kind of lamer doing that code13:40
@lisitsyn:D13:40
@besser82lisitsyn, Remeber me to smack you on next SG_Workshop  :P13:41
@lisitsynright13:41
@besser82lisitsyn, recompiling...  just a few minutes13:44
@besser82lisitsyn, problem is not double free'ing...13:59
@besser82lisitsyn, must be some other problem...14:00
@besser82lisitsyn, let's wait for the build-job finishing14:00
@besser82lisitsyn, hah!14:05
@besser82lisitsyn, do we have some option / method in SG to determine whether the ptr has already been unref()'ed?14:06
@lisitsynbesser82: what ptr?14:07
@lisitsynsgobject or sgmatrix?14:07
@besser82lisitsyn, CDynamicObjectArray*  :P14:08
@besser82lisitsyn, so should be sgobject, rye?14:09
@lisitsynyes14:09
@besser82lisitsyn, CDynamicObjectArray* m_binary_evaluations get's unref'ed by calling swig's `_wrap_delete_CrossValidationMulticlassStorage`14:10
@lisitsynyes14:10
@lisitsynm_binary_evaluations14:10
@lisitsynline 6014:10
@besser82lisitsyn, yes...  and later we are free'ing the unref'ed shiatz...  :P14:11
@besser82*BOOM*14:11
@besser82brb in ~ 15 minutes..., lisitsyn14:11
@lisitsynwhy?14:11
@lisitsynI don't get where do we free unrefed thing14:13
-!- nszceta [~nszceta@5ED05E57.cm-7-1b.dynamic.ziggo.nl] has joined #shogun14:31
@besser82lisitsyn, re14:31
@besser82lisitsyn, _wrap_delete_CrossValidationMulticlassStorage calls shogun::CSGObject::unref on CDynamicObjectArray* m_binary_evaluations14:32
@besser82lisitsyn, and later on the dtor of CrossValidationMulticlassStorage14:34
@besser82lisitsyn, http://fpaste.org/207198/14:37
@besser82lisitsyn, the whole class get's unref'ed14:37
@besser82lisitsyn, better paste: http://fpaste.org/207199/81510451/14:38
@besser82lisitsyn, complete func:  http://fpaste.org/207200/51190142/14:40
-!- nszceta [~nszceta@5ED05E57.cm-7-1b.dynamic.ziggo.nl] has quit [Remote host closed the connection]15:01
-!- nszceta [~nszceta@5ED05E57.cm-7-1b.dynamic.ziggo.nl] has joined #shogun15:02
-!- nszceta_ [~nszceta@95.211.152.2] has joined #shogun15:06
-!- nszceta [~nszceta@5ED05E57.cm-7-1b.dynamic.ziggo.nl] has quit [Ping timeout: 256 seconds]15:09
@besser82lisitsyn, ?15:09
@besser82lisitsyn, are you dead, man?15:09
@besser82lisitsyn, I found it, I think...15:10
@lisitsynbesser82: I atent dead!15:18
@besser82lisitsyn, I see :P15:18
@lisitsynok back15:18
@besser82lisitsyn, so CSGObject::unref() invokes `delete this`15:18
@lisitsynhmm yes15:18
@lisitsynso?15:18
@besser82lisitsyn, so SG_UNREF(Object*) frees the whol object15:19
@lisitsynyou're talking about that storage object, right?15:19
@besser82lisitsyn, nope15:19
@besser82lisitsyn, or yes15:19
@besser82lisitsyn, the object swig unrefs15:20
@lisitsynhmm okay15:20
@besser82lisitsyn, and that looks like the whole class to me15:20
@lisitsynbesser82: I am not sure I get the problem15:23
@besser82lisitsyn, the problem is: the shiats blows-up when calling the destructor...15:24
@lisitsynyes15:24
@besser82lisitsyn, or immediatly after...15:25
@besser82lisitsyn, lemme check which of the two deletes is causing the blow...15:30
@lisitsynbesser82: I don't get the two deletes thing15:30
@lisitsynwhy there are two delets?15:31
@besser82lisitsyn, whether the delete[] m_fold_ROC_graphs15:31
@besser82lisitsyn, or the UNREF-delete of the class itself...15:31
@besser82lisitsyn, done by swig15:32
@besser82lisitsyn, it's definetly ../src/shogun/evaluation/CrossValidationMulticlassStorage.cpp:4115:33
@besser82lisitsyn, SG_FREE(m_fold_ROC_graphs);15:33
@besser82lisitsyn, and that blows15:33
@lisitsynbesser82: any idea why swig interferes with that?15:34
@besser82lisitsyn, no real idea...  :(15:35
@lisitsynso you commented out15:35
@lisitsyn4115:35
@lisitsynand it works?15:36
@besser82lisitsyn, yes15:36
@lisitsynbesser82: did you try to check whether m_fold_ROC_graphs is not null at line 36?15:38
@besser82lisitsyn, yes15:38
@lisitsynstill fails?15:38
@lisitsynunless you comment free?15:38
@besser82lisitsyn, yes15:38
@besser82lisitsyn, and reproducible with 2.0.1215:39
@besser82lisitsyn, so it's genereal problem now15:39
@besser82*general15:39
@besser82reproducible wit hswig 2.0.1215:40
@besser82lisitsyn, ^  ;(15:40
@lisitsynok I see15:40
@besser82lisitsyn, I try direct delete[] on that line now15:40
@besser82lisitsyn, let's see what happenes then :P15:41
-!- kshitijk [~kshitijk@220.227.149.86] has quit [Ping timeout: 265 seconds]15:43
@lisitsynbesser82: delete[] would kaboom15:44
@besser82lisitsyn, yes seen now too  :P15:44
@lisitsynyou can't delete[] what is malloc15:44
@besser82lisitsyn, dammit, yes!15:44
@besser82lisitsyn, so that's the reason for da boom :P15:44
@lisitsynwhy?15:45
@besser82lisitsyn, you just said it...  ;)15:45
@lisitsynbut SG_FREE is free15:45
@lisitsynand SG_MALLOC is malloc15:45
@besser82lisitsyn, nope15:45
@lisitsynI am looking at code in memory.cpp and it is ;)15:46
@besser82template<> void sg_generic_free<type >(type* ptr)                                                      $15:46
@besser82{                                                                                                      $15:46
@besser82        delete[] ptr;                                                                                  $15:46
@besser82}15:46
@lisitsynoh15:46
@lisitsynreally?15:46
@besser82yes!15:46
@besser82dammit15:46
@besser82there was da shaitz cooking @15:46
@lisitsynoh15:47
@lisitsyngosh15:47
@besser82lisitsyn, me too  :P15:47
@lisitsynso we have specialized mallocs15:47
@lisitsynfor sgmatrices15:47
@besser82lisitsyn, yes15:47
@besser82lisitsyn, but no free()'s15:47
@besser82lisitsyn, lemme fix that15:47
@besser82lisitsyn, as it always and only happenes with SGMatrix15:48
-!- besser82 [~besser82@fedora/besser82] has left #shogun ["Verlassend"]15:48
-!- besser82 [~besser82@fedora/besser82] has joined #shogun15:48
-!- mode/#shogun [+o besser82] by ChanServ15:49
@lisitsynbesser82: you can kill lines 38,39,46,47,54,55,73,74,81,82,93,9415:50
@lisitsynas they are called with new[] and delete[] anyway15:50
@besser82lisitsyn, no replacement?15:50
@besser82lisitsyn, ok I see ^^15:51
@lisitsynthese in plcae news and explicit destructor calls were made for case15:51
@lisitsynwhen malloc/free is used15:51
@lisitsynbut at some point someone introduced these specialized mallocs..15:51
@lisitsynI didn't even knew about that :D15:51
@besser82lisitsyn, leeme do test run...15:52
@lisitsynI think this should fixi t15:52
@besser82lisitsyn, commented lines 38,39 and uncommented the SG_FREE()15:53
@besser82lisitsyn, let's see where it goes...15:53
@lisitsynbesser82: 46,47 could fail then15:53
@lisitsyn:D15:53
@besser82lisitsyn, nope... that shit is untested  :P15:59
@besser82lisitsyn, will apply same sscheme to evaluation_multiclassovrevaluation_modular15:59
-!- kshitijk [~kshitijk@220.227.149.86] has joined #shogun16:04
-!- PirosB3 [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has quit [Quit: PirosB3]16:06
@lisitsynbesser82: so this fixes shiatz or?16:07
@besser82lisitsyn, yes, shit now works, no segfault  :D16:07
@lisitsynI see16:07
@besser82lisitsyn, so delete[] tried to recursively kill all stuff in ptr16:08
@lisitsynyes16:08
@lisitsyntried to delete already deleted shiatz16:08
@besser82lisitsyn, but stuff was already delete[]'ed :D16:08
@besser82lisitsyn, and that goes *BOOM*16:08
@lisitsynKAWABANGA16:08
@lisitsynbesser82: btw older swig doesn't work on my machine asw ell16:11
@lisitsynas well*16:11
@lisitsynits yosemite shitz16:11
@besser82lisitsyn, so apple fsked it up  :D16:11
@lisitsynbesser82: oh16:15
@lisitsynI think I know the answer16:15
@lisitsynfscking shitze16:15
@lisitsynbesser82: I think cmake detected wrong python16:16
@lisitsynbesser82: you probably can help with that16:16
-!- kshitijk [~kshitijk@220.227.149.86] has quit [Ping timeout: 248 seconds]16:18
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun16:24
shogun-notifier-shogun: Bj?rn Esser :bugfix/python_modular_swig3 * 2b6c88d / src/shogun/evaluation/ (2 files): https://github.com/shogun-toolbox/shogun/commit/2b6c88d7ff2e94a83d8b75593090100764c828e716:24
shogun-notifier-shogun: don't call delete[] twice on SGMatrix-instances16:24
@besser82lisitsyn, gimme output from cmake plz ---> fpaste.org16:25
shogun-notifier-shogun: Bj?rn Esser :develop * 2067632 / examples/undocumented/python_modular/ (2 files): https://github.com/shogun-toolbox/shogun/commit/2067632151fee2c60d239a2c4dbc4fdcfb34ae6d16:28
shogun-notifier-shogun: fix failing test when building with swig 3.X16:28
shogun-notifier-shogun: Bj?rn Esser :develop * 2b6c88d / src/shogun/evaluation/ (2 files): https://github.com/shogun-toolbox/shogun/commit/2b6c88d7ff2e94a83d8b75593090100764c828e716:28
shogun-notifier-shogun: don't call delete[] twice on SGMatrix-instances16:28
shogun-notifier-shogun: Bj?rn Esser :develop * 203d5c3 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/203d5c305127cb0a9dc8cd773063b792f76e30f116:28
shogun-notifier-shogun: Merge pull request #2698 from shogun-toolbox/bugfix/python_modular_swig316:28
shogun-notifier-shogun:16:28
shogun-notifier-shogun: fix segfaults when calling delete[] twice on SGMatrix-instances and fix minor issues with swig 3.X16:28
@lisitsynbesser82: do you know how to make cmake to output what python it found?16:30
@besser82lisitsyn, it should show that in messages displayed...16:31
@besser82lisitsyn, just paste and I'll try to figure out :P16:31
@lisitsynbesser82: I think I have some interesting lines16:31
@lisitsynbesser82: http://fpaste.org/207220/42815795/16:33
@lisitsynI think this thing is totally messed16:33
@lisitsynI have python installed with homebrew16:33
shogun-buildbotbuild #267 of trusty - libshogun - viennacl is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/267  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>16:33
@besser82lisitsyn, I see...16:34
@besser82lisitsyn, and that is detected, but not the correct libs  :(16:34
@besser82lisitsyn, lemme think a few minutes...16:35
@besser82lisitsyn, which python do you want to use?16:36
@besser82lisitsyn, system's or homebrew's?16:36
@lisitsynbesser82: I think we should use the one that is called with 'python'16:36
@lisitsynin command line16:36
@besser82lisitsyn, allrighty...  homebrews16:37
@lisitsynthere must be some consistency16:37
@lisitsynyou know :)16:37
@lisitsynit is easy to find a fix for that16:38
@lisitsynI mean workaround16:38
@lisitsynbut for users we need a real fix16:38
-!- travis-ci [~travis-ci@ec2-54-89-18-89.compute-1.amazonaws.com] has joined #shogun16:38
travis-ciit's Bj?rn Esser's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/5715760516:38
-!- travis-ci [~travis-ci@ec2-54-89-18-89.compute-1.amazonaws.com] has left #shogun []16:38
@besser82lisitsyn, cmake -DPYTHON_INCLUDE_DIR:PATH=/usr/local/include/python${X.X} -DPYTHON_LIBRARY:FILEPATH=/usr/local/lib/libpython${X.X}.dylib -DPYTHON_PACKAGES_PATH:PATH=/usr/local/lib/python${X.X}/site-packages16:39
@besser82lisitsyn, replace ${X.X} with real version  :P16:39
@besser82lisitsyn, that's not so easy to fix that in real16:39
@besser82lisitsyn, because that would require a hell of checking...16:40
@lisitsynbesser82: I don't have /usr/local/include/python16:41
@lisitsynit is in /usr/local/Cellar/python16:41
@lisitsynso this could be easier16:41
@besser82lisitsyn, then use this one  ;)16:41
@besser82lisitsyn, nope...  because of libs16:41
@besser82lisitsyn, cmake takes the first libs found, which match the {major, minor}-version...16:42
@besser82lisitsyn, and my bindings goes *BOOM* :P16:42
-!- PirosB3 [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has joined #shogun16:43
@lisitsynbesser82: okay I adjusted this thing manually16:45
@lisitsynlets see if it fails16:46
@lisitsynI think it won't16:46
@besser82lisitsyn, I think all should work ^^16:48
@besser82lisitsyn, build with "patched" swig 3  and run the tests, plz  ;)16:48
@lisitsynbesser82: I can finally kaggle16:50
@lisitsyn:D16:50
@lisitsynI hope16:50
@besser82lisitsyn, hehe  :D16:50
@lisitsynbesser82: ja ja ja16:53
@lisitsynja16:53
@lisitsyn>>> import modshogun16:54
@lisitsyn>>> modshogun.KMeans()16:54
@lisitsynKMeans16:54
@besser82lisitsyn, strike!16:58
@besser82lisitsyn, python_modular-tests?  with swig3?  :P16:58
@lisitsynchecking17:01
-!- travis-ci [~travis-ci@ec2-54-197-139-24.compute-1.amazonaws.com] has joined #shogun17:03
travis-ciit's Bj?rn Esser's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/5715790417:03
-!- travis-ci [~travis-ci@ec2-54-197-139-24.compute-1.amazonaws.com] has left #shogun []17:03
@lisitsynbesser82: like 99% works17:09
@lisitsynapart from libraries like hdf5 etc17:10
@lisitsynthey are just missed so that's ok17:10
@lisitsynbesser82: bingo :)17:10
@besser82lisitsyn, Yay!  =)17:10
@besser82lisitsyn, do you have an knowledge about json-c`17:10
@besser82?17:10
@lisitsynnot really17:11
@besser82lisitsyn, damn :(17:11
@lisitsynwhat's it you want to know?17:11
@besser82lisitsyn, json-c 0.12 fscked up basically json-serializion in sg :(17:11
@lisitsynheh interesting17:12
@besser82lisitsyn, http://fpaste.org/207224/81603901/17:13
@besser82lisitsyn, UAIFileTest is fscked-up, too --->  http://fpaste.org/207225/81604661/17:15
shogun-notifier-shogun: Bj?rn Esser :develop * 3adb24b / .gitignore: https://github.com/shogun-toolbox/shogun/commit/3adb24bbe5e1a6e5221b473cf479acd4bf9fd99817:21
shogun-notifier-shogun: updated gitignore17:21
shogun-buildbotbuild #268 of trusty - libshogun - viennacl is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/268  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>17:23
-!- kshitijk [~kshitijk@220.227.149.86] has joined #shogun17:35
-!- travis-ci [~travis-ci@ec2-54-197-139-24.compute-1.amazonaws.com] has joined #shogun17:41
travis-ciit's Bj?rn Esser's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/5716269417:41
-!- travis-ci [~travis-ci@ec2-54-197-139-24.compute-1.amazonaws.com] has left #shogun []17:41
-!- HeikoS [~heiko@dab-ntm1-h-55-3.dab.02.net] has joined #shogun18:00
-!- mode/#shogun [+o HeikoS] by ChanServ18:00
@besser82HeikoS, Hey!18:06
-!- nszceta_ [~nszceta@95.211.152.2] has quit [Quit: My Mac has gone to sleep. ZZZzzz...]18:07
-!- PirosB3 [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has quit [Quit: PirosB3]18:07
@HeikoSbesser82: hi!18:17
@HeikoSlisitsyn: hi!18:17
@HeikoSlisitsyn: just checking out the manual stuff, pretty cool that it works now18:17
@lisitsynHeikoS: yaeap18:18
@besser82HeikoS, finally fixed the problem with SWIG3 :D18:18
@HeikoSbesser82: really? amazing!18:18
@HeikoSbesser82: saw the PR but didnt yet look in detail18:18
@lisitsynHeikoS: yeah and I found the reason18:18
@HeikoSbut cool18:18
@HeikoSput it in the news18:18
@lisitsynwhy mac fails18:18
@besser82HeikoS, https://github.com/swig/swig/pull/372  ;)18:18
@HeikoSlisitsyn: why was it?18:18
@lisitsynHeikoS: cmake linking to wrong python18:19
@besser82HeikoS, the PR in Shogun was just minor stuff18:19
@HeikoSlisitsyn: ah cool18:19
@HeikoSbesser82, lisitsyn amazing guys, this means mac is now working finally?18:19
@HeikoSthe way to macports and homebrew is ready then :)18:19
@lisitsynHeikoS: there is a workaround to fix that18:19
@lisitsynbut we didn't fix it yet18:19
-!- PirosB3 [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has joined #shogun18:20
@besser82HeikoS, real fix need some real deep work into CMake stuff...18:20
@lisitsynbut this seems to be easier than to understand why it crashed18:20
@HeikoSbesser82: ok?18:20
@HeikoSlisitsyn: cool18:20
@HeikoSgood progress then18:20
@HeikoSamazingf18:20
@HeikoSlisitsyn: so about the manual stuff18:21
@lisitsynI have hard times remembering - crazy errors in debugger - check whether the library is correct18:21
@HeikoSthe tabs are global now18:21
@lisitsynyes?18:22
@HeikoSlisitsyn: can we only have them in examples?18:22
@besser82HeikoS, I can look into it...  but that might take some time18:22
@lisitsynyeah probbaly18:22
@HeikoSor you think they should be everywhere?18:22
@HeikoSI mean eventually, all examples will be in one folder18:22
@HeikoSso we can just check that and only add tabs if needed18:22
@lisitsynthat's adjustable with template I guess18:23
@HeikoSlet me check18:23
@HeikoSlisitsyn: ah yeah of course18:23
@HeikoSIll do it18:23
@lisitsynHeikoS: if you find a way to parametrize what languages we support18:24
@lisitsynthen it would be prettier18:24
@HeikoShow do you mena?18:24
@HeikoSlike not hard-coded?18:24
@lisitsynyeah18:24
@HeikoSwe can just easily load this from a text file18:24
@HeikoSand create it with cmake?18:24
@lisitsynI don't know how to pass them into jinja18:24
@lisitsynno that's too much18:24
@HeikoShow you wanna do it?18:24
@lisitsynpassing them to jinja is enough18:24
@lisitsynbut I didn't found a way to pass a list to jinja there18:25
@HeikoSI see18:26
@HeikoSwhat about just building the string you pass?18:26
@lisitsynHeikoS: building what string?18:26
@HeikoSlisitsyn: I guess I did not see what you meant18:27
@HeikoSpassing to jinja=?18:27
@HeikoSwhere?18:27
@lisitsynHeikoS: ok let me explain the problem18:27
@lisitsynin template we have all these divs for python octave java18:27
@HeikoS<script>18:28
@HeikoS$('#python-tab').on('show.bs.tab', function (e) {18:28
@HeikoS    $('ul[id^=tabs] a[href^=#python]').tab('show');18:28
@HeikoS});18:28
@HeikoSthis stuff?18:28
@lisitsynand in our sgexample extension we have get_supported_languages18:28
@lisitsynyes18:28
@lisitsynif we template that18:28
@lisitsynwould be better18:28
@lisitsynlike #language-tab18:28
@HeikoSI see18:28
@HeikoSI agree18:28
@HeikoSbut not too bad for now?18:28
@HeikoSas languages do not really change severely18:28
@lisitsynyeah18:28
@HeikoSlisitsyn: do you know how to do if-then-else in template?18:29
@HeikoSthen we could mark the example pages somehow18:29
@HeikoSso that page.html only is active if example page?18:29
@lisitsyniirc it is like if endif18:29
@HeikoSlisitsyn: ok18:30
@HeikoSlisitsyn: and how to check whether page is example?18:30
@HeikoScan one like check from which folder it comes?18:30
@HeikoSor mark the .rst somwhow?18:30
@lisitsynhmm don't know18:30
@lisitsynthere are some globals18:30
@lisitsynhttp://sphinx-doc.org/templating.html18:31
@lisitsynHeikoS: looks like should be possible to do via that stuff18:32
@HeikoSyeah should be18:33
@HeikoSnbut still dont know how to mark18:33
@lisitsynnot clear for me too18:33
@HeikoS{% if example %}18:33
@HeikoS{% endif %}18:33
@HeikoSand all the stuff in between18:33
@HeikoSbut how to define exam?ple?18:34
@lisitsynyeap18:34
@lisitsynhmm18:34
@HeikoShttp://stackoverflow.com/questions/13209597/override-html-page-template-for-a-specific-sphinx-document18:34
@HeikoSpagename18:34
@HeikoSmay18:34
@HeikoSbut thats "knn" here18:35
@lisitsynHeikoS: then we should force names18:35
@lisitsyn:)18:35
@HeikoSpagename (relative doc path of each file)18:35
@HeikoSlisitsyn: mmmg18:36
@HeikoSthat path does it right=?18:36
@HeikoSif pagename.split(os.sep)[1] == "examples"18:37
@lisitsynhmm18:37
@lisitsynshould work?18:37
@HeikoSlisitsyn: trying18:40
@HeikoSlisitsyn: any idea on this:18:47
@HeikoSlisitsyn: we currently have in the .rst file:18:47
@HeikoS.. sgexample:: ../code/knn.sg18:47
@HeikoS  :language: python18:47
@HeikoS  :start-after: ![load_data]18:47
@HeikoS  :end-before: ![load_data]18:47
@HeikoSbut we really want:18:47
@HeikoS.. sgexample:: knn.sg18:47
@HeikoSsorry18:47
@HeikoSwe want18:47
@HeikoS\snippet knn load_data18:47
@lisitsynHeikoS: what's this 'language' for?18:47
@HeikoSlisitsyn: useless18:47
@lisitsynok18:48
@HeikoSjust need it to overwrite it later18:48
@HeikoSmaybe do something before sphinx comes in?18:48
@HeikoSor also do it with sphinx?18:48
@HeikoSalso it would be good to have a command for referecing a shogun class18:48
@HeikoS(this should be easier)18:48
@HeikoSyou know, just to make the .rst files shorter and nicer looking18:48
@lisitsynHeikoS: I am not sure I get the idea18:56
@HeikoSlisitsyn: check the rst file18:56
@lisitsynHeikoS: okay language can be removed18:57
@lisitsynnow you want to remove start-after18:57
@lisitsynand end-before18:58
@lisitsynright?18:58
@HeikoSthere are like 5 lines18:59
@HeikoSwould be cool to have just 118:59
@lisitsynokay19:00
@lisitsyn.. sgexample:: examples/knn.sg load_data19:01
@lisitsynHeikoS: like that?19:01
@HeikoSlisitsyn: yeah19:02
@HeikoSlisitsyn: really much cleaner19:03
@HeikoSand also19:03
@lisitsynok let me do that19:03
@HeikoS.. sgclass CEuclideanDistance19:03
@HeikoSlisitsyn: once we have that, we are basically ready to structure the thing a bit19:03
@HeikoSand put a build process around it19:03
@HeikoSthat translates examples etc19:03
@HeikoSlisitsyn: ok got the tabs working for example pages19:07
@HeikoSlisitsyn: pushed19:12
@lisitsyncool19:12
@lisitsynHeikoS: I am trying to get this thing to work19:12
@lisitsynone liner19:12
@lisitsynI broke start-before thing somehow :D19:12
@lisitsynah19:13
@lisitsynstart-after19:13
@lisitsyn:D19:13
@lisitsynokie19:14
@lisitsyndone19:14
@HeikoShaha19:16
@HeikoSpushed?19:16
@lisitsynnot yet19:16
@lisitsynin a minute19:16
@HeikoSgogogog :)19:16
@HeikoSlisitsyn: btw will you join these video meetings?19:16
@lisitsynHeikoS: if available19:16
@HeikoSI did this with gunnar a few times and it really helps focussing19:16
@HeikoSgoal is still this hackathon and funding for a software engineer19:16
@HeikoSwe are applying to an nsf grant atm19:17
@HeikoSlisitsyn: but anyways19:17
@HeikoSsend the push ;)19:17
@lisitsyn`HeikoS19:17
@lisitsyndone19:18
@lisitsynoneliner now19:18
-!- PirosB3 [~pirosb3@host33-99-dynamic.181-80-r.retail.telecomitalia.it] has quit [Quit: PirosB3]19:18
@HeikoSlisitsyn: one thing I realise19:18
@HeikoSlisitsyn: do you have this whitespace on the knn example html page?19:18
@HeikoSthere is like a margin when the tabs are visible19:18
@HeikoSlisitsyn: ah beautiful with the one-liner19:19
@HeikoSlisitsyn: how hard is an ..sgclass ?19:19
@lisitsynhmm nothing strange I see19:19
@lisitsynwhat should sgclass do?19:19
@HeikoSlisitsyn: link to19:19
@HeikoShttp://www.shogun-toolbox.org/doc/en/latest/19:19
@HeikoSlink to the class API doc19:19
@lisitsynI think we should capture shogun classes automatically19:19
@HeikoSyeah exactly so that people dont paste links19:20
@lisitsynboth in text and code19:20
@HeikoSas the link location might change19:20
@HeikoSah you mean19:20
@HeikoSwe just write19:20
@lisitsynyeah19:20
@HeikoS"CKNN" and then its done automagically?19:20
@lisitsynand javascript19:20
@HeikoScool19:20
@HeikoSeven better so the .rst is super clean then19:20
@lisitsynshould be possible19:20
@HeikoShow to do that?19:20
@lisitsynhmm let me think19:21
@HeikoSlisitsyn: I mean scraping would be best19:23
@lisitsynHeikoS: I'll think about it19:23
@HeikoSso any C++ Shogun name should be extracted19:23
@HeikoSalso the enums19:23
@lisitsynyes19:23
@lisitsynand methods19:23
@lisitsyn:D19:23
@HeikoSyes19:23
@lisitsynmethods is harder19:23
@HeikoSlisitsyn: we could just scrape19:23
@lisitsynbut should be possible19:23
@HeikoSCKNN::train19:23
@HeikoSso that the context will be known19:24
@lisitsynHeikoS: ok I'll do that a bit later19:24
@lisitsynHeikoS: I think I know good way to do that19:24
@HeikoSkk19:24
@lisitsynHeikoS: in sphinx I will add some code to glance over all text19:24
@HeikoSlisitsyn: cool19:24
@lisitsynand replace C*19:24
@HeikoSlisitsyn: so what else is missing?19:24
@HeikoSplots?19:25
@lisitsynwith <p id="sg3423423">C*</p>19:25
@HeikoSlisitsyn: scikit learn has these cool plots19:25
@HeikoSlisitsyn: shall we just not do them and rather link to notebook?19:25
@lisitsynthen $('p[id^=sg])') and so on19:25
@HeikoSThink that might be better19:25
@HeikoSlisitsyn: yeah got it19:25
@lisitsynHeikoS: i think we should overshogun sklearn19:25
@lisitsynI want web plots19:25
@HeikoSlisitsyn: hehe19:25
@HeikoSlisitsyn: so how to do that?19:25
@lisitsynpossible19:25
@HeikoSlisitsyn: keep in mind examples are extracted from meta language19:26
@lisitsynhmm19:26
@lisitsynplots are done from python19:26
@lisitsyn;)19:26
@HeikoSyeah19:26
@lisitsynthat's it19:26
@HeikoSbut where is the code coming from?19:26
@lisitsynnot get it19:26
shogun-notifier-shogun: Wu Lin :develop * 33ae686 / / (7 files): https://github.com/shogun-toolbox/shogun/commit/33ae68616cc7578c4b95de9772dbbe118eb0f96119:26
shogun-notifier-shogun: update the ARD kernels19:26
shogun-notifier-shogun: Wu Lin :develop * 2c8c39e / src/shogun/kernel/GaussianARDKernel.h: https://github.com/shogun-toolbox/shogun/commit/2c8c39ec01bfccedd93cdea827c9046a899e156f19:26
shogun-notifier-shogun: fix a typo19:26
shogun-notifier-shogun: Wu Lin :develop * 15e36ea / tests/unit/kernel/GaussianARDKernel_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/15e36ea0fbeda3606b19df07dda26ef7288483ba19:26
shogun-notifier-shogun: add another use case19:26
shogun-notifier-shogun: Heiko Strathmann :develop * 071ab37 / / (7 files): https://github.com/shogun-toolbox/shogun/commit/071ab37a055820f2eb77e7904e01688568f261fa19:26
shogun-notifier-shogun: Merge pull request #2796 from yorkerlin/ardkernel19:26
shogun-notifier-shogun:19:26
shogun-notifier-shogun: update the method for ARD kernels19:26
@HeikoSlisitsyn: btw wu is writing amazing code these days19:27
@HeikoSlisitsyn: could we also have a wikipedia link thing?19:28
@HeikoSlisitsyn: do you think that is a good idea?19:28
@HeikoSor weblink in general?19:28
@HeikoSthese are hard-links I guess?19:28
@lisitsynHeikoS: wikipedia? where to?19:28
@HeikoSlisitsyn: like for more reference on the algorihtm19:28
@HeikoSwe should have a unified way to provide further reading19:28
@HeikoSlike19:28
@HeikoSwiki, papers, scikit-.learn docs etc19:29
@lisitsynah19:29
@HeikoSshould be under "references" at bottom of page19:29
@HeikoSbut automatically created19:29
@lisitsynyeah we can do that as well19:29
@HeikoSso that we can change the way this looks19:29
@HeikoSwithout changing all files19:29
@lisitsynyes sure19:29
@HeikoSI dont want to touch this ever again, everything should be done automatically ;)19:29
@HeikoSlazy19:29
@lisitsynthis is good19:29
@lisitsynHeikoS: well19:30
@lisitsynshould be easy to do19:30
@lisitsynI think I have some understanding of these directives now19:30
@HeikoSlisitsyn: btw do you have this whitespace margin issue?19:30
@HeikoSlisitsyn: not whitespace, margin19:30
@lisitsyncan you show me screenshot?19:30
@HeikoSlisitsyn: a good idea would also be to document the code a bit more19:30
@HeikoSyeah coming19:30
@HeikoSlisitsyn: mail sent19:31
shogun-buildbotbuild #269 of trusty - libshogun - viennacl is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/269  blamelist: Wu Lin <yorker.lin@gmail.com>19:31
@lisitsynwhat margin you're talking about?19:31
@lisitsynleft/right?19:31
shogun-buildbotbuild #270 of trusty - libshogun - viennacl is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/270  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Wu Lin <yorker.lin@gmail.com>19:32
@lisitsynthere is no margin like that you have here19:32
@lisitsynbut code snippets are too wide for some reason19:32
@HeikoSmmh19:33
@HeikoSthere is white space between the content of the example page and the menu on the left19:33
-!- nszceta [~nszceta@95.211.152.2] has joined #shogun19:33
@lisitsynyeah19:33
@lisitsynstrange19:33
@HeikoSlisitsyn: this is not there at the mainpage19:33
@HeikoSgoes away19:33
@lisitsyncause by tabs then19:33
@lisitsyncaused*19:33
@HeikoSyeah19:34
@HeikoSlisitsyn: do you think we should a a quick documentation about the sphinx plugin?19:35
@lisitsynHeikoS: don't know19:35
@lisitsynprobably19:35
@lisitsynand more error checking19:36
@lisitsynsorry have to go19:36
@lisitsynI'll check these issues a bit later19:36
@HeikoSlisitsyn:  ok bye19:37
-!- travis-ci [~travis-ci@ec2-54-197-139-24.compute-1.amazonaws.com] has joined #shogun19:52
travis-ciit's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/5717300319:52
-!- travis-ci [~travis-ci@ec2-54-197-139-24.compute-1.amazonaws.com] has left #shogun []19:52
-!- nszceta [~nszceta@95.211.152.2] has quit [Quit: My Mac has gone to sleep. ZZZzzz...]19:52
-!- curiousguy13 [~curiousgu@122.162.129.17] has joined #shogun21:13
-!- HeikoS [~heiko@dab-ntm1-h-55-3.dab.02.net] has quit [Quit: Leaving.]21:31
@besser82lisitsyn, wiking: yo?21:51
@besser82lisitsyn, wiking: I have a small question / discussion...21:51
@lisitsynbesser82: whats up21:52
@besser82lisitsyn, Would be fine with me updating the requirement on json-c to 0.12?21:52
@besser82lisitsyn, there have been a bunch of changes in api...21:52
@lisitsynbesser82: what's version in ubuntu 14.04?21:52
@besser82lisitsyn, 0.11  :(21:53
-!- pickle27 [~pickle27@192-0-134-235.cpe.teksavvy.com] has joined #shogun21:53
@besser82lisitsyn, but who cares about ubuntu?21:53
@besser82lisitsyn, Fedora rulez da world!  =)21:53
@lisitsynbesser82: hmm can we link statically?21:53
@besser82lisitsyn, yes, we can21:54
@lisitsyncan we bundle it?21:54
@lisitsynlike download sources21:54
@besser82lisitsyn, -DBUNDLE_JSON=ON  :P21:54
@besser82lisitsyn, option is in cmake21:54
@lisitsynbuild and do static link21:54
@besser82lisitsyn, and Fedora Engineering will kick my ass...21:55
@lisitsynain't good?21:55
@besser82lisitsyn, forbidden items:21:55
@besser82lisitsyn, 1. BUNDLED LIBS21:55
@besser82lisitsyn, 2. STATIC LINKING21:55
@lisitsynwe need an option then21:55
@besser82lisitsyn, what option?21:56
@lisitsynlike do normal way21:56
@lisitsynor forbidden :)21:56
@besser82lisitsyn, with normal way I need update req. to 0.1221:56
@besser82lisitsyn, and rewrite some parts of SerializationJSON21:56
@besser82lisitsyn, to use the new api21:57
@besser82lisitsyn, and shogun will be future proof =)21:57
@besser82lisitsyn, so what shall we do?21:58
@besser82lisitsyn, upgrade shogun stuff?21:58
@lisitsynhmm I am ok with updating probably21:58
@besser82lisitsyn, we need upgrade :P21:59
@lisitsynwell it would be best21:59
@lisitsynto make it work with zero changes on ubuntu and mac21:59
@besser82lisitsyn, Fedora is the only dist with real fully featured shogun ^^21:59
@lisitsynthese are majorities by user21:59
@lisitsyns21:59
@besser82lisitsyn, then bring them to fedora  ;)22:00
@besser82lisitsyn, What Ubuntu?22:00
@lisitsynlast lts22:00
@besser82lisitsyn, Use Fedora22:00
@besser82lisitsyn, Fully feature shogun out of box + ML-Community  ;)22:00
@besser82lisitsyn, that's not with ubuntu-crap :P22:00
@lisitsynI don't want any further restrictions as we already nearly impossible to install :)22:01
@besser82lisitsyn, are we?22:01
@besser82lisitsyn, on Fedora all looks fine but JSON  :P22:01
@lisitsynyeah users struggle quite a lot22:01
@besser82lisitsyn, They need to use Fedora :P22:01
@besser82lisitsyn, sudo yum install shogun22:01
@besser82lisitsyn, sudo yum install ${lang}-shogun22:01
@besser82lisitsyn, done22:01
@lisitsynwe need sudo apt-get install shogun22:02
@lisitsynand brew install shogun22:02
@besser82lisitsyn, you can doooooooooooooooo it!  :D22:02
@lisitsynunless we have it working22:02
@lisitsynit sucks22:02
@besser82lisitsyn, I probably can start working on ppa22:02
@besser82lisitsyn, end of next week or so22:02
@lisitsynbesser82: sounds cool22:03
@besser82lisitsyn, so we have "sudo apt-get install shogun" on the most lame OS ever, just before MacOSX  :P22:04
@besser82lupinix, jetzt sag doch auch mal was :P22:04
lupinixtach besser82 :P22:05
* lupinix votes for fedora22:05
@besser82lupinix, about why it's the best to have shogun on Fedora exclusive  ^^22:05
@besser82lisitsyn, ^ see?22:06
lupinixbesser82: don't forget redhat enterprise/centos ;)22:06
@besser82lupinix, WIP ^^22:06
@lisitsynsomeone on RHEL?22:06
@besser82lupinix, For RHEL / CentOS / SL 5+  :D22:06
@besser82lisitsyn, Not only RHEL... CentOS, SL and Oracle, too22:07
@besser82lisitsyn, so we cover all premium OS around...  ;)22:07
@lisitsynpremium haha22:08
@lisitsynlol22:08
@besser82lisitsyn, why lol?22:15
@besser82lisitsyn, ok... Oracle is crap22:15
@lisitsynI have no idea who's using that22:15
@lisitsyn:D22:15
@besser82lisitsyn, but the others are pretty good22:15
@besser82lisitsyn, people like me?22:15
@besser82lisitsyn, amazon...22:16
@besser82lisitsyn, cisco22:16
@besser82lisitsyn, SAP22:16
@besser82lisitsyn, other $big_money_corporates22:16
@lisitsynheh22:16
@lisitsynokaokay22:16
@besser82lisitsyn, if we have the rpms, we can get easy into those and collect da $$$22:17
@besser82lisitsyn, for the SHOGUN YMIOP ltd.22:17
@besser82YMIOP == Your Money In Our Pockets22:19
@besser82lisitsyn, ^22:19
@lisitsynhaha I see22:19
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]22:27
-!- HeikoS [~heiko@05453a2b.skybroadband.com] has joined #shogun22:38
-!- mode/#shogun [+o HeikoS] by ChanServ22:38
@lisitsynHeikoS: hey22:40
@HeikoSlisitsyn: hey!22:40
@HeikoSnice mail22:40
@HeikoSneeded22:40
@lisitsynHeikoS: need to optimize multiclass classification for log loss22:40
@lisitsynneed to predict probabilities22:41
@HeikoSlisitsyn: aha?22:41
@lisitsynsuggestions? :)22:41
@HeikoSlisitsyn: dont have too much time right now, wanna send me an email?22:41
@HeikoSlisitsyn: there should be known methods for that?22:41
@lisitsynHeikoS: ah sorry22:41
@lisitsynin shogun22:41
@lisitsynlets talk later then22:41
@HeikoSlisitsyn: no worries ,just went online to check something, but now leaving again for dinner22:41
@HeikoSlisitsyn: sure!22:41
@HeikoSlisitsyn: get back to me on this, i am interested!22:42
@lisitsynHeikoS: kaggle ;)22:42
-!- HeikoS [~heiko@05453a2b.skybroadband.com] has quit [Client Quit]22:43
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 265 seconds]23:37
-!- HeikoS [~heiko@05453a2b.skybroadband.com] has joined #shogun23:55
-!- mode/#shogun [+o HeikoS] by ChanServ23:55
--- Log closed Sun Apr 05 00:00:29 2015
« Friday, 2015-04-03 Index Sunday, 2015-04-05 »

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