Open in new window / Try shogun cloud
--- Log opened Thu Aug 08 00:00:06 2013
-!- travis-ci [] has joined #shogun00:00
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []00:00
-!- lisitsyn [] has quit [Quit: Leaving.]00:34
-!- van51 [] has joined #shogun00:58
-!- van51 [] has left #shogun ["PING 1375919736"]01:55
-!- shogun-notifier- [] has quit [Quit: transmission timeout]02:02
-!- zxtx [] has quit [Ping timeout: 246 seconds]04:27
-!- gsomix [~gsomix@] has joined #shogun06:24
@sonney2kgsomix, moin moin!06:27
gsomixsonney2k, moin. makin' changes, preparin' PR.06:28
gsomixsonney2k, what time you begin working day? I mean in TomTom.06:29
-!- shogun-notifier- [] has joined #shogun08:17
shogun-notifier-shogun: Soeren Sonnenburg :develop * efe9a9f / examples/undocumented/python_modular/ (6 files):
shogun-notifier-shogun: convert some more kernel examples to CSVFile08:17
shogun-buildbotbuild #1682 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Soeren Sonnenburg <>08:23
sonne|workgsomix: now :)08:36
gsomixsonne|work, huh, ok.08:37
gsomixsonne|work, I changed CSVFile to double reading instead twice memory usage08:38
sonne|workcool but tests are more important to fix08:38
sonne|workgsomix: so fix tests -> PR08:39
sonne|workthen do the rest08:39
-!- travis-ci [] has joined #shogun08:47
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []08:47
-!- zxtx [] has joined #shogun09:10
gsomixsonne|work, there?09:25
gsomixsonne|work, err
gsomixI finally got confused.09:39
sonne|workgsomix: which of the files should I look at?09:41
gsomixsonney2k, second09:42
-!- van51 [] has joined #shogun09:42
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:60ed:c7da:bf1a:7110] has joined #shogun09:43
-!- mode/#shogun [+o iglesiasg] by ChanServ09:43
@iglesiasggood morning, good morning09:43
gsomixiglesiasg, good? okay. :)09:44
@iglesiasggsomix, not good for you?09:44
gsomixiglesiasg, I'd said chaotic neutral.09:45
shogun-notifier-shogun: Viktor Gal :feature/CMake * e827cd1 / CMakeLists.txt,src/shogun/lib/,src/shogun/mathematics/lapack.h:
shogun-notifier-shogun: Add proper Accelerate framework detection09:49
@wikingso who loves the smell of napalm in the morning? he could maybe merge ...?09:50
-!- lisitsyn [] has joined #shogun09:51
@iglesiasgwiking, you should have the pleasure of merging it!09:53
sonne|workgsomix: yes exactly like this09:53
gsomixsonney2k, \o/09:54
sonne|worklisitsyn: so you are now officially from a terrorist country?09:54
lisitsynsonne|work: what's it?09:55
lisitsynI am not following your media ;)09:55
lisitsynsonne|work: ahh snowden case09:57
sonne|workcold war II is coming09:58
lisitsynno, that's not true09:58
sonne|workor in other words winter is coming09:58
@iglesiasgbrace yourselves09:58
lisitsynjust games, in the long run US is a pure friend because all the stolen russian money is there09:59
lisitsynsonne|work: as for snowden I do not really understand what made the people so surprised10:01
@iglesiasghello hello10:04
lisitsyn"Your mother's so fat, she's everyone's nearest neighbor."10:05
thoralflisitsyn: lolo10:05
lisitsyniglesiasg: large mother nearest neighbor10:05
@iglesiasgLMNN indeed10:05
thoralflisitsyn: You're an agent provocateur to make us saying bad things about the us? ;)10:06
lisitsyniglesiasg: the method is to drop the first nearest neighbor (the mother) that adds bias10:06
lisitsynthoralf: haha are you disallowed to do that?10:06
thoralflisitsyn: No, we're free to do that.10:06
lisitsynI am not actually I like the most of the things the US do10:07
thoralflisitsyn: But since it's in the channel logs, maybe my next travel so sf will be rejected? ;)10:07
thoralfFree to do anything.  (With all consequences ;))10:07
lisitsynthoralf: I hope to get visa soon too!10:08
thoralflisitsyn: Then say something bad about Putin&Snowden&Co.10:08
thoralflisitsyn: Now.10:08
lisitsynthoralf: I am totally apathetic to snowden actually10:09
thoralflisitsyn: Or better: Refuse to say something political.  You're just a normal guy that wants to work and carry its money to the US. :)10:09
thoralflisitsyn: Very good. :)10:09
lisitsynbut the putin is fucker we are desperate to get rid of him10:09
lisitsynI don't know anyone with IQ > 70 that supports him10:10
lisitsynwe have to live with it though10:11
lisitsynthoralf: no I just want to visit mentor summit ;)10:11
sonne|worklisitsyn: I guess you passed the via us entrance test10:12
thoralflisitsyn: I think Putin is very clever (and reckless) and a good networker within the "old" russian executive. ;)10:12
lisitsynthoralf: no he is totally mad nowadyas10:12
lisitsynhe was like clever and good in 200410:12
lisitsynsonne|work: well I participated in rallys in 2011 ;) so that's what I think10:13
thoralflisitsyn: But how can you explain that he's still your leader?  Seems that he's not legitimated by the people, so there is another reason.10:13
lisitsynthoralf: there is a team that supports him10:13
sonne|workthoralf: well why do we have merkel?10:14
thoralfsonne|work: Don't remind me on that.10:14
lisitsynthoralf: russian people is mostly stupid - they would prefer to keep things10:14
sonne|worklisitsyn: that holds for all people10:14
lisitsynI'd say he is supported by 30-40% now10:14
lisitsynthat's the exact rate of idiots10:15
lisitsynthoralf: is merkel that bad?10:15
lisitsynI have no clue10:15
lisitsynbut you have a chance to get rid of she, right?10:16
thoralflisitsyn: Yes10:16
lisitsynour guy is eternal10:16
thoralflisitsyn: No10:16
thoralflisitsyn: Next election won't stop her.10:17
lisitsynthoralf: how?10:17
-!- travis-ci [] has joined #shogun10:17
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []10:17
lisitsynthoralf: isn't it already second time or so?10:17
lisitsynI mean isn't it limited?10:17
thoralflisitsyn: No, unlimited.10:17
lisitsynlive with it10:17
thoralflisitsyn: 4-years * #successful-elections ;)10:18
lisitsynbut do you think elections are cheated?10:18
lisitsynor she is just supported by the majority10:18
thoralflisitsyn: No.10:18
thoralfNo cheating, I guess.10:18
lisitsynthoralf: here they are ;)10:18
lisitsynit was heaaavily cheated10:18
lisitsynthoralf: but what do you want then? that's how democracy works10:19
thoralflisitsyn: There's no difference between cheating elections and doing-what-your-lobby-wants-once-elected. ;)10:19
lisitsynand that's why democracy is bad too10:19
lisitsynthoralf: I think there is10:19
thoralflisitsyn: Yes, I not fully serious.10:20
sonne|workguys stop whining10:20
sonne|workpeople are evil10:20
lisitsynsonne|work: no, why whining10:20
sonne|workonly shogun can rescue them :)10:20
sonne|workso start coding!10:20
lisitsynI just note what is wrong ;)10:21
thoralfYes, only the tough guys are getting the cool chicks.10:21
thoralfSo start beeing tough!  Now!10:21
lisitsynokay let me write my MoveOperationsIntersectionChainValidator10:21
lisitsynthoralf: MEAT10:22
lisitsynthoralf: CHICKS10:22
lisitsynthoralf: WEAPONS10:22
sonne|worklisitsyn: well you have the first 210:22
thoralfPanem et circenses :)10:22
sonne|worklisitsyn: if you count shogun as weapon also 310:22
lisitsynsonne|work: CARS10:22
lisitsynARGHH MADNESS!10:23
lisitsynthoralf: is it tough enough?10:23
thoralfsonne|work: Having "gun" in the name doesn't make it a weapon. :)10:23
lisitsynI just realized gunnar could be spelled like gunner10:23
thoralflisitsyn: I don't know.  I'm no cool chick. :)10:23
lisitsynlike machinegunner or whatever10:23
lisitsynthoralf: do you know any?10:24
van51sonne|work: hello10:24
thoralflisitsyn: "All of them"[tm]10:24
van51sonne|work: can you take a look at a comment I just made on the PR?10:24
thoralfBtw., gotta work now more serious.  See you.10:24
shogun-notifier-shogun: Viktor Gal :feature/CMake * dfa109e / tests/unit/io/
shogun-notifier-shogun: Fix LineReader_unittest with cmake10:25
@wikinggsomix: next time use the __FILE__ macro..10:26
lisitsynwiking: are you ready to merge it?10:26
@wikinglisitsyn: i think so10:26
@wikinglisitsyn: i'm waiting now this output10:26
@wikingi mean travis... since most of the things are going to get green10:26
@wikingbut let's see10:27
lisitsynI see10:27
@wikingi've got the weirdest error on the osx job (not on my machine locally though..):
lisitsynwiking: so he believes std::complex<float> is incomplete10:30
sonne|workvan51: url?10:31
sonne|workvan51: wait you just do while (true) {     CDynamicArray* hashed_indices = new CDynamicArray();10:36
sonne|workSG_UNREF(hashed_indices); }10:36
@wikinglisitsyn: ah ok i've got this with my compiler as well... heh i think i've fucked up the things with -framework Accelerate10:36
sonne|workvan51: and that gives you the error?10:36
van51sonne|work: it's not an infinite loop but yeah10:37
van51sonne|work: the error comes when called from the apply method of a multiclass svm10:37
van51sonne|work: with num_threads > 110:37
sonne|workvan51: with c++11 enabled?10:39
van51sonne|work: yea I think it's enabled10:40
@wikingheh yey two greens on travis \o/10:40
-!- foulwall [~user@] has joined #shogun10:42
sonne|workvan51: please edit lib/config.h and .config and remove the HAVE_CXX11 and try if the problem still exists10:43
van51sonne|work: ok10:43
-!- travis-ci [] has joined #shogun10:54
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []10:54
thoralfsonne|work: If you've got 2 minutes, I've got a design question: Why do you always check dimensions when doing dot products etc.?  When (for example) multiplying two vectors, one could assume that it's infinite and filled with zeros.11:08
thoralfsonne|work: Makes things lot more complicated when streaming stuff.11:08
thoralfsonne|work: I usually don't know how many dimensions the final data will have after transformation, so it would be nice if it could be a bit more tolerant.11:09
sonne|workthoralf: reason is just to avoid errors - we had a couple of kaboom situations where this did help11:09
sonne|workthoralf: but sure for streamign I get your point11:11
thoralfsonne|work: Do you see a way to relax this?11:12
sonne|workthoralf: well you could do yourself - just give it vectors with adjusted size11:12
thoralfsonne|work: I mean, assuming infinite vectors that are implicit filled with zeros is no big deal (except that .set_const() won't be clearly defined)11:12
thoralfsonne|work: This would be the same (semantically), yes.  Just truncating inputs until they fit. ;)11:14
sonne|workthoralf: yeah but it misses the error case - and it would be easy to adjust the size of your vector11:15
thoralfsonne|work: But it's not clear how to "truncate" sparse vectors without copying then.11:15
sonne|workthoralf: I mean x=y; x.vlen=17;11:15
sonne|workthoralf: ahh sparse11:16
-!- gsomix [~gsomix@] has quit [Ping timeout: 264 seconds]11:20
thoralfsonne|work: Actually it doesn't matter if sparse or dense: The problem that training-inputs have lower dimensions than evaluating-inputs makes things complicated.  Either: more checks/expensive resizing/reading data twice11:21
van51sonne|work: I hope I removed it correctly. I removed what you said plus some compiler flags mentioning c++11 and it is still happening11:21
-!- gsomix [~gsomix@] has joined #shogun11:21
thoralfsonne|work: But i acknowledge that I might not overlook all the implications of changing this behavious.11:22
sonne|workthoralf: well for dense one has to do d=min(d1,d2) and then call dot with appropriate size11:22
sonne|workthoralf: for sparse dims are IIRC not checked11:23
sonne|workthoralf: indeed not checked11:23
sonne|workthoralf: which function do you mean11:23
sonne|workvan51: alright then we have a nice race condition I cannot currently make sense of - but you did call init_shogun_with_defaults() once right?11:24
thoralfsonne|work: for example CLinearMachine::apply_get_outputs11:25
van51sonne|work: yeah11:25
van51sonne|work: this is the entire example
thoralfsonne|work: But seen this in other places, too, since I'm using sparse all the time.11:26
van51sonne|work: I'll continue to look into it11:26
sonne|workvan51: please create a minimal unit test for that11:26
van51sonne|work: ok11:28
sonne|workthoralf: yeah but it is not in SGSparseVector sparse_dot at least11:28
sonne|workgsomix: tests fixed?11:29
hushelliglesiasg: hi, do you have time to check my PR?11:30
@iglesiasghushell, let mesee11:30
@iglesiasghushell, did you update it already?11:31
hushelliglesiasg: yes, few hours ago, when you sleeping I guess11:31
gsomixsonne|work, fixing.11:31
thoralfsonne|work: Right, but actually it doesn't matter where to relax things.  Just make it work more robust, since the product can be computed naturally without this checks.11:31
thoralfsonne|work: Any suggestions?11:31
@iglesiasghushell, it is a shame I don't get mails when new commits are done. If you remember for the next time, just write a dummy comment when you make a new commit, something like, updated! :)11:32
@iglesiasghushell, I will check anyway if I can modify my github settings to get notified more11:32
thoralfsonne|work: Maybe just remove the redundant checks where it hurts sparse?11:32
sonne|workthoralf:  it will be various places. if possible you should do it in a way that it still helps catching erros11:32
thoralfsonne|work: DenseFeatures seem to duplicate these checks.11:32
sonne|workthoralf: it makes sense for streaming to remove checks11:33
sonne|workbut not for full matrices IMHO11:33
sonne|workmaybe for sparse it does11:33
sonne|workbecause you never know how much dims you will get when testing11:33
hushelliglesiasg: really? seems github isn't that reliable11:36
thoralfsonne|work: I'll check what I can do with minimal impact. ;)11:36
sonne|workiglesiasg: you sure you subscribed to getting these notifications?11:36
@iglesiasgsonne|work, my notification status is watching11:37
sonne|workthoralf: I guess you are right with sparse but you need to make the code then more tolerant as in when a sparse vector with dim > dim_dense arrives drop indices11:37
sonne|workhmmh no clue11:37
hushellThe time tag for the new commit is always wrong11:37
hushellonly records the first commit11:38
@iglesiasgsonne|work, I actually get mails if anyone comments something in a pull request I have commented. But for some reason I don't get any mail if only a new commits happens or someone amends a commit11:38
@iglesiasghushell, might that be because of the timezone?11:39
hushelliglesiasg: I don't think so, your comments are right11:39
hushellI meant the time11:40
@iglesiasghushell, I think it is looking ready! There is just this question about reference counting of m_fg.11:45
hushelliglesiasg: I just answered your question :)11:50
@iglesiasghushell, thank you!11:52
-!- gsomix [~gsomix@] has quit [Ping timeout: 264 seconds]11:56
@iglesiasgI just read about kmeans++, do we have that already in Shogun? Otherwise, I will write an issue. It looks like a nice entrance task.12:01
@iglesiasgIt seems we don't have it12:06
thoralfsonne|work: It basically affects CSparseFeatures.  If you want to keep your checks, what about having a class "RelaxedSparseFeatures", which with as less dimension-assumptions as possible and SparseFeatures, which just overloads and adds assertions?12:11
sonne|workiglesiasg: what is kmeans++?12:12
@iglesiasgsonne|work, an extension of kmeans to initialize in a smart way the cluster centers12:12
sonne|workthoralf: it will also make dense_dot in sparse expensive since you have to check for each dim12:12
sonne|workiglesiasg: so 'just' different cluster center init?12:12
@iglesiasgsonne|work, pretty straightforward to understand:
-!- gsomix [~gsomix@] has joined #shogun12:13
-!- gsomix [~gsomix@] has quit [Client Quit]12:13
@iglesiasgsonne|work, yeah, although 'only this' seems to improve kmeans quite a bit12:13
thoralfsonne|work: Having these checks doesn't affect performance in my case.  (Less than 1%)12:13
-!- gsomix [~gsomix@] has joined #shogun12:13
sonne|workthoralf: well no when you take a sparse vector you have to check for each dim if it is < valid dim12:14
thoralfsonne|work: I know.  In the loop.12:14
sonne|workthoralf: so it will affect everything12:15
gsomixsonne|work, hey12:17
gsomixsonne|work, fixed. but, please, check it twice.12:18
thoralfsonne|work: No big deal.  I can leave it as it is and patch my local installation.  No risk breaking something upstream.12:18
-!- van51 [] has quit [Ping timeout: 260 seconds]12:23
@iglesiasghushell, everything working fine in the unit tests after the updates?12:23
@iglesiasghushell, are ./shogun-unit-test --gtest_filter=FactorGraph* and valgrind ./shogun-unit-test --gtest_filter=FactorGraph* happy?12:24
hushellFactorGraph* happy but there is a CSV test failed12:24
@iglesiasgthat's fine, gsomix will take care of CSV ;)12:25
hushelliglesiasg: I didn't add unit tests for this PR, since it's just framework code12:25
@iglesiasghushell, yep. But there were unit tests in the one before testing the disjoint set stuff, right?12:26
@iglesiasgI meant it because of those12:26
@iglesiasgsince disjoint set was refactored here12:26
hushelliglesiasg: yeah, there are good after these changes12:26
@iglesiasghushell, valgrind too? :)12:26
hushelliglesiasg: let me recheck the valgrind12:27
@iglesiasghushell, thank you12:27
hushelliglesiasg: yep, they are fine as well12:28
@iglesiasghushell, awesome, here we go then12:28
hushelliglesiasg: shall we wait the travis?12:28
@iglesiasghushell, not really. It will take long since we have these tests that hang up due to old eigen in travis12:29
shogun-notifier-shogun: hushell :develop * 5902995 / src/shogun/structure/ (6 files):
shogun-notifier-shogun: MAP Inference for factor graph12:29
shogun-notifier-shogun: Fernando Iglesias :develop * 3b61d41 / src/shogun/structure/ (6 files):
shogun-notifier-shogun: Merge pull request #1376 from hushell/fg_dev12:29
shogun-notifier-shogun: MAP Inference for factor graph12:29
hushelliglesiasg: well, I'll go to prepare a new PR12:29
@iglesiasghushell, some method for inference already?12:30
hushelliglesiasg: just max product :( I need make sure the SSVM happy with the factor graph model first12:31
@iglesiasghushell, it is cool anyway. What SSVM will you be using? PrimalMosekSOSVM?12:31
hushelliglesiasg: I'll try the primal one first, since you know better about this12:34
@iglesiasghushell, yep. For bundle methods like BMRM, they either work out of the box, or we will need to ask Michal12:34
shogun-buildbotbuild #1684 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Fernando Iglesias <>12:34
hushelliglesiasg: but I'd like to have a demo like Michal's face landmark detection12:38
@iglesiasghushell, that would be really cool, though challenging12:38
@wikingwhat's happening with [ RUN      ] GaussianDistribution.sample_2d12:39
@iglesiasgwiking, the old eigen issue12:40
@wikingiglesiasg: 3.0.5?12:40
@iglesiasgwiking, it has been like that for some days now. That test never finishes in old eigen12:40
@iglesiasgwiking, the one in ubuntu repositories12:40
@wikingsoon soon12:40
@iglesiasgwiking, me touching travis.yaml the other day was an attempt to fix that ;)12:40
@wikingjust trying to find out wtf is with ruby modular12:41
@wikingit crashes on my machine.... let's hope that it's some osx bs12:41
@wikingsince the other day it was still working...12:41
@iglesiasglet's see12:41
@wikingotherwise i'll move the ruby job to allowed to fail ;)12:42
@wikingiglesiasg: i've put 3.1.2 for eigen12:42
@wikingas a minimum requirement in cmake12:42
@wikingthat should do it, right?12:42
@iglesiasgyes, I think so12:42
@iglesiasglet me double check with the version I installed12:42
@iglesiasgit was 3.1.4 IIRC12:43
@wikingok otherwise if one wants to have bundled eigen then it'll download 3.1.412:43
@wikingyeah but requiring 3.1.4 is quite unrealistic12:43
@wikingas it's really couple of days old12:43
@iglesiasgI think Heiko said 3.1 should be fine12:43
@iglesiasgI think ubuntu ships some 3.0. something12:43
@wikingnext ubuntu version (12.10 or 13.x) has 3.1.212:43
@wikingit's just that travis has 12.0412:44
@wikingand that has 3.012:44
thoralfDoes shogun implement a SVM solver which can maximize F-score?12:52
-!- HeikoS [] has joined #shogun13:01
-!- mode/#shogun [+o HeikoS] by ChanServ13:01
gsomixsonne|work, ok, there are some issues, should fix.13:09
gsomixsonne|work, don't merge commit.13:10
gsomixcu little later13:12
-!- gsomix [~gsomix@] has quit [Remote host closed the connection]13:17
@wikingHeikoS: yo man13:18
@wikingHeikoS: so i was just about to ask yesterday that13:18
@wikingis that script working for ya?13:18
@HeikoSwiking yes13:18
@wikingto generate the ipython notebooks13:18
@wikingbecause for me it's like this13:18
@HeikoSwiking: but you need ipython-1.013:18
@HeikoSwhich you can only get from git13:19
@HeikoSyeah I know imports13:19
@wikingi see13:19
@wikingso it's only ipython 1.0 ready13:19
@wikingokok so we need that13:20
@wikingno worries13:20
shogun-buildbotbuild #1683 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: hushell <hushell@hushell-U510.(none)>13:22
thoralfIs it allowed that a sparse feature vector contains two entries for the same feature index?  CSparseFeatures<ST>::get_feature(num,index) suggests something, but is this intended?13:22
@HeikoSwiking: can you tell cmake to get ipython 1.0?13:27
@HeikoSthoralf: I think this causes problems13:27
@HeikoSthoralf: for example when converting to a dense, I know that the results will be random13:27
shogun-notifier-shogun: Viktor Gal :feature/CMake * b300468 / CMakeLists.txt,src/shogun/mathematics/lapack.h:
shogun-notifier-shogun: Fix Accelerate framework with c++11 and linking13:28
shogun-notifier-shogun: Viktor Gal :feature/CMake * a911bd8 / / (4 files):
shogun-notifier-shogun: Disable c++11 mode if using libc++ and compiling modular interface13:28
shogun-notifier-shogun: Fix ruby_modular library naming and test environment13:28
shogun-notifier-shogun: Remove unused CMAKE_C_FLAGS from external gmock project13:28
@wikingHeikoS: heheh lets see13:29
-!- foulwall [~user@] has quit [Read error: Connection reset by peer]13:30
-!- foulwall [~user@] has joined #shogun13:30
-!- foulwall [~user@] has quit [Read error: Connection reset by peer]13:32
@wikingwould be good enough?13:33
@wikingHeikoS: this is insanely huge13:34
-!- foulwall [~user@] has joined #shogun13:34
@wikingHeikoS: imho we dont need this in cmake....13:37
@wikingHeikoS: i mean you can run the ipynb-s with the 0.1313:37
@wikingonly to run this script we need 1.013:37
@wikingand that we'll take care ourselves at a buildbot13:37
@wikingand if somebody wants to run it as well13:38
@wikinghe/she will download/install ipython 1.0 himself13:38
-!- foulwall` [~user@] has joined #shogun13:39
-!- foulwall [~user@] has quit [Ping timeout: 246 seconds]13:40
@HeikoSwiking: agreed13:44
@HeikoSyeah we just download a copy ourselves13:47
@HeikoSusers dont need to run and upload the notebooks13:47
@HeikoSthey just can just open them13:48
@wikingok lets see what does this build do finally13:51
@HeikoSwiking: I like the allowed failures!13:52
@HeikoSproper hacking ;)13:52
@wikinghas InferenceMethod.get_log_ml_estimate_binary_logit_laplace been hacked lately?14:07
@wikingmeaning within the last 48 hours14:07
@HeikoSwiking: its random so it might fail14:07
@wikingmmm it fails sometimes14:07
@HeikoSdecrease accuracy in this case14:07
@HeikoSjust by 20% or so14:07
@wikingok do it14:08
@wikingu know what u r talking about ;)14:08
@HeikoSok wil ldo14:08
@wikinghehe it's 43 degrees outside \o/14:12
@HeikoSwiking: whaat?14:13
@HeikoSwhere are you?14:13
-!- gsomix [~gsomix@] has joined #shogun14:16
gsomixok, let's check python tests14:16
-!- travis-ci [] has joined #shogun14:17
travis-ci[travis-ci] it's Fernando Iglesias's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []14:17
@wikingHeikoS: remote reporting14:21
hushelliglesiasg: I sent a new PR :) please have a look. talk you tomorrow14:21
@iglesiasghushell, yes, just saw it. Thank you! See you then14:22
hushellgood, CU14:22
-!- hushell [~hushell@] has quit [Quit: WeeChat 0.3.7]14:22
@HeikoSwiking: cool!14:22
sonne|workHeikoS: UTC?14:22
sonne|workthoralf: not worth the hassle14:23
@HeikoSwiking: and much faster14:23
@HeikoSsonne|work: doodle tell you the timezone14:23
@HeikoSit shows me mine14:23
@HeikoSsonne|work: but the one in the evening today is 22 german time14:23
@iglesiasgwiking, that looks fancy!14:23
@HeikoSsonne|work: the bank account thing keeps on being annyoing14:24
@iglesiasgHeikoS, it actually shows here just 9:00 PM so it is good that you mentioned it!14:24
@HeikoSsonne|work: it *is* set up now, but it was a few days too late and now my parents are in vaccation, so noone can tell me the number14:24
@HeikoSsonne|work: I called them but they refuse to give it over phone14:24
sonne|workHeikoS: it didn't set me the the time zone14:24
@HeikoSiglesiasg: whatever doodle days14:24
sonne|workHeikoS: when are they back?14:25
@HeikoSsonne|work: weeks14:25
@HeikoSsonne|work: so there are a few options now:14:25
sonne|workHeikoS: as in still agust?14:25
@HeikoSsonne|work: yes end of august14:25
@iglesiasgHeikoS, yep but the thing is 9:00 your timezone or ours? :P It14:25
@HeikoS1.) do it over my girl's account (no income in gemrnay)14:25
sonne|workthen it should be fine14:25
@HeikoS2.) ask my grandparent to find the mail (unlikely)14:25
sonne|workHeikoS: in any case we have to do the whole supplier enrolment again14:26
sonne|workI guess we can still wait.14:26
@HeikoSsonne|work: I know but I need a bank account14:26
sonne|workHeikoS: do you already have your account number and everything?14:26
@HeikoSsonne|work: no thats the problem14:26
@HeikoSsonne|work: they wont tell me wihtout the pin which is in a letter at home14:26
sonne|workHeikoS: so you don't even  know your 'kontonummer' ?14:26
sonne|workwe dont' need the pin now14:27
@HeikoSI know14:27
@HeikoSbut they dont give it14:27
@HeikoSvia phone14:27
@wikingdo u need an offshore acc?14:27
sonne|workHeikoS: then I guess we better wait14:27
@HeikoSwhen is the PO deadline?14:27
sonne|workHeikoS: now I still don't know which timezone doodle is in14:28
@HeikoSsonne|work: doodle shows the local time14:28
sonne|workbut I am not available this evening anyways14:28
@HeikoSah ok14:28
sonne|workso better just do it with the others in the evening and then we could have a chat tomorrow in the morning14:28
@HeikoSso fill in the doodle14:29
@HeikoSwe can also wait till next week its not urgent14:29
@HeikoSoh btw I am away next weel14:29
@HeikoSI think everyone should be there14:29
sonne|workHeikoS: well doc camp?14:29
@HeikoSsonne|work: what does the thing show for you? which time options?14:30
sonne|work10 am 4pm 9pm14:30
-!- foulwall` [~user@] has quit [Read error: Connection reset by peer]14:30
@HeikoSah ok14:30
@HeikoSthats local then14:30
@HeikoSwait no14:30
@HeikoSshows the same here14:30
@HeikoSwhat time is it in germany?14:30
@HeikoSwow then doodle does not show local14:31
@HeikoSsorry for this14:31
@HeikoSso doodle is utc +114:31
sonne|workHeikoS: anyway most pressing in the moment is doc camp yes / no14:32
@HeikoSsonne|work: lets reschedule the meeting then, in two weeks or so14:32
-!- foulwall` [~user@] has joined #shogun14:32
@HeikoSsonne|work: doc camp I would say yes14:32
sonne|workcan you write sth?14:32
@HeikoSI got some time today and tomorrow to work on the proposal14:32
@HeikoSsonne|work: yeah my parents are back end of august14:33
sonne|workbut maybe you have a quick meeting with the others and you all do your votes14:33
@HeikoSso thats fine14:33
@wikingso i guess no:)14:33
@HeikoSWe will have a meeting this evening then without you sonne|work then we can discuss14:33
@HeikoSand you can edit things tomorrow14:33
@wikingsonne|work: did u make the octave_modular interface?14:35
sonne|workwiking yes sure14:35
-!- foulwall` [~user@] has quit [Read error: Connection reset by peer]14:35
@wikingfor i in $(SWIGLIBFILES) ; do echo $$i | sed -e 's=.*/==' -e 's/\.oct/;/' \ >>$(DESTDIR)$(OCTDIR)/shogun/init_shogun.m; done14:36
@wikingwhere what and how?14:36
thoralfsonne|work: thoralf: not worth the hassle <-- Talking about what?14:36
thoralfHeikoS: It's only random because some places are dealing with duplicated features, others don't.14:37
@HeikoSthoralf: yes14:37
@HeikoSbut no idea about the order etc14:37
@HeikoSalso saving etc14:37
@HeikoSyou can probably store multiple ones, but no defined behaviour afterwards14:37
thoralfHeikoS: Before doing anyhing - what is the desired behaviour?14:38
thoralfHeikoS: Even duplicated features are okay - but only if it's consistent. :)14:39
@wikingsonne|work: i dont know how to generate that init_shogun.m14:39
sonne|workwiking: just all modules14:40
thoralfHeikoS, sonne|work: I'll send a PR which makes duplicated features consistent, okay?14:40
@HeikoSthoralf: sounds good!14:40
@wikingsonne|work: ?14:40
@HeikoSthoralf: why do you need that even?14:40
sonne|workthoralf: optimize the f-score!14:41
thoralfHeikoS: First of all: When hashing features, you get duplicates.14:41
sonne|workthoralf: it is not helping / so joachims paper from icml long time ago14:41
thoralfsonne|work: Do you mean SVMperf?14:41
sonne|workthoralf: don't remember sth that can optimzie F1/ auROC etc14:42
thoralfsonne|work: I guess it's ICML 2005, svm perf.14:43
thoralfsonne|work: But there is nothing comparable implemented in shogun, right?14:43
@wikingthoralf: u use octave?14:44
thoralfHaving 3M input vectors and only 2k positives.  That's bad thing for accuracy. ;)14:44
thoralfwiking: Actually not, but I could.  Why?14:44
@wikingi need the content of init_shogun.m14:45
thoralfwiking: Lol.  Second.14:45
thoralfnitialize shogun by loading all of its modules14:45
thoralfThats it.14:45
@wikingmmm interesting14:45
@wikingthen creating14:46
gsomixsonne|work, hey14:47
thoralfwiking: Have a look at ./src/interfaces/octave_static/Makefile:14:47
sonne|workthoralf: no but it is not helping performance anyway so not worth the time14:48
sonne|workwiking: so drop it please14:48
sonne|workwiking: before we had like 10 modules Kernel, Distance etc etc14:49
sonne|worksince we now only have one it doesn't make sense any longer14:49
@wikingsonne|work: you mean to drop the init_shogun.m ?14:49
sonne|workwiking: yes14:50
@wikingsonne|work:  imean the problem is then that the octave examples need to be changed14:50
sonne|workfor i in *.m ; do sed -i s/init_shogun/modshogun/g $i ; done14:50
sonne|workwiking: ^14:50
gsomixsonne|work, can you help me with python example? look, fm_test_real contains 120 lines, fm_train_real - 92 lines.14:52
-!- van51 [] has joined #shogun14:52
gsomixsonne|work, kernel.init(feats_train, feats_test)14:52
gsomixSystemError: [ERROR] train or test features #dimension mismatch (l:92 vs. r:120)14:52
gsomixwhat's wrong? data? example? reader?14:52
sonne|workgsomix: look at the python examples I converted14:52
sonne|workgsomix: e.g. gaussian_kernel_modular.py14:52
van51sonne|work: the first test case here : is a minimal example of that segfault14:53
gsomixsonne|work, File "", line 16, in kernel_gaussian_modular14:53
gsomix    kernel.init(feats_train, feats_test)14:53
gsomixSystemError: [ERROR] train or test features #dimension mismatch (l:92 vs. r:120)14:53
sonne|workvan51: first test case is?14:54
van51sonne|work: crash_test14:54
sonne|workgsomix: then you did break it again14:54
gsomixsonne|work, how?14:54
gsomixsonne|work, my tests works fine14:55
sonne|workgsomix: how what? This works here.14:55
sonne|workgsomix: did you git pull?14:55
gsomixsonne|work, of course.14:55
sonne|workvan51: hmmhh too long to figure it out14:55
van51sonne|work: here :
gsomixsonne|work, but I did changes in CSVFile a little.14:55
-!- travis-ci [] has joined #shogun14:56
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []14:56
van51sonne|work: in the dense_dot method, if I remove these news and unref's it doesn't happen14:56
sonne|workgsomix: so this is the script right ?14:56
sonne|workgsomix: this works so I guess you've broken it in your CSVFile upate14:57
gsomixsonne|work, yep, I have this script. latest version.14:58
gsomixsonne|work, aww.14:58
sonne|workgsomix: ok then your fault(tm)14:58
gsomix:( I don't know what we need then.14:58
sonne|workgsomix: what's the diff?14:58
sonne|workgsomix: well tests were all broken14:58
sonne|workgsomix: did you fix the test with the *original* version?14:58
sonne|workof CSVFile?14:58
gsomixsonne|work, yes. but my tests works.14:59
sonne|workvan51: hmmhh not really a minimal example :/14:59
gsomixsonne|work, yep. I created new branch from latest develop14:59
gsomixI changed a little CVSFile14:59
sonne|workvan51: does it fail with Multiclass ocas too?14:59
gsomixbecause it was working wrong14:59
sonne|workit was working fine14:59
van51sonne|work: let me check15:00
sonne|workvan51: one more question15:00
sonne|workvan51: if you add init_shogun_with_defaults() as first line in this test - does it work?15:00
gsomixsonne|work, please, look again.
gsomixnow CSVFile works in this way.15:01
sonne|workgsomix: bit it did already work like this15:02
van51sonne|work: yes it still fails if I add that function call15:02
gsomixok, let me check15:02
thoralfsonne|work: Our discussion about sparse vectors and their length etc. is almost obsolete: Someone already fixed it in develop ;)15:07
shogun-notifier-shogun: Viktor Gal :feature/CMake * cdf52bf / / (81 files):
shogun-notifier-shogun: Dropping init_shogun.m in octave_modular interface15:07
shogun-notifier-shogun: Viktor Gal :feature/CMake * 0c3449d / src/interfaces/octave_modular/CMakeLists.txt:
shogun-notifier-shogun: remove lib prefix for octave_modular dylib15:07
shogun-notifier-shogun: Viktor Gal :feature/CMake * efd0ba6 / examples/undocumented/ruby_modular/CMakeLists.txt:
shogun-notifier-shogun: Add narray libary's path to the test env15:07
sonne|workthoralf: what?15:07
thoralfsonne|work: So bounds checking in the dot-product-loops is already done - it's now very easy to eliminate the assertions.15:08
thoralfsonne|work: 2 weeks ago.15:08
sonne|workquite some slow down15:09
thoralfsonne|work: No!15:09
thoralfsonne|work: Didn't affect performance at all.15:09
sonne|workwhy not? is branch prediction that good?15:09
thoralfsonne|work: Just ran two versions.15:09
thoralfI don't know.15:09
gsomixsonne|work, can you help me a little?15:22
sonne|workgsomix: how?15:25
gsomixhow many vectors and features?15:26
gsomixthere are 2 rows and 5 cols.15:26
sonne|workgsomix: 2 features, 5 vectors15:26
gsomixsonne|work, noooop15:26
gsomixso rows==num_feats15:27
gsomixin matrix that I read there are 2 rows and 5 cols.15:27
lisitsynsonne|work: I see it as 2 vectors too :D15:27
gsomixand matrix = [1, 2, 3, 4, 5, 6, 7, 8, 9, 0]15:28
sonne|workgsomix: don't mix up shogun representations with ascii15:28
sonne|worklisitsyn: you must be blind15:28
sonne|worklisitsyn: one line == one vector15:28
lisitsynsonne|work: I don't know what format you are talking about15:28
lisitsynbut if it was a matrix15:28
sonne|worklisitsyn: csv15:28
lisitsyna feature matrix15:28
sonne|workit is not15:29
lisitsynohkay I must be blind ;)15:29
sonne|workgsomix: but yes num_feat == num_rows (in an SGMatrix not in a CSV file!)15:30
sonne|workgsomix: OK?15:31
van51sonne|work: what about this example?
van51sonne|work: can you also test it?15:43
sonne|workvan51: no I cannot test anything currently15:43
sonne|workvan51: but you forgot the pthread_join !15:44
sonne|workvan51: maybe that is the reason it crashes in the multiclass code?15:44
shogun-notifier-shogun: Viktor Gal :feature/CMake * 5108fab / .travis.yml,src/interfaces/octave_modular/CMakeLists.txt:
shogun-notifier-shogun: Fix typo on in octave_modular interface cmake file15:45
shogun-notifier-shogun: install numpy with pip in python 3.3 travis env15:45
van51sonne|work: dunno, not really that experienced with threads :)15:45
sonne|workvan51: well when you create threads you have to wait until they are finished15:45
sonne|workvan51: that is what pthread_join does15:45
sonne|workotherwise you might have some still running15:45
van51sonne|work: ok, let me add it and see15:46
sonne|workjust a loop before exit_shogun15:46
van51sonne|work: a loop?15:48
sonne|workvan51: well for each thready you create you have to call pthread_join - you want them all to finish right?15:49
van51sonne|work: aah ok15:49
van51sonne|work: I though you meant to keep calling join for the same thread15:49
van51sonne|work: hmm it doesn't happen that often15:50
sonne|workvan51: no :)15:50
gsomixsonne|work, it seems I figured out finally15:52
gsomixI'm little crazy with all rows, cols, feats and vecs.15:53
sonne|workgsomix: yeah it is tough15:53
-!- naywhayare [] has joined #shogun15:57
gsomixlisitsyn, engineering is fun16:01
thoralfsonne|work: You've been removing sort_features() from CSparseFeatures?16:02
thoralfsonne|work: I need it in sparseVector, mind if I move it and call it from sparse matrix in a loop?16:02
sonne|workthoralf: it should be in SGSparseMatrix now16:03
thoralfsonne|work: It is.  But only in a loop.16:03
thoralfsonne|work: I need it for single vectors.16:03
sonne|workerrm IIRC we never had one for single vectors16:04
thoralfsonne|work: iirc we had - in sparse features.16:04
thoralfsonne|work: No, you're right.16:06
-!- travis-ci [] has joined #shogun16:11
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []16:11
gsomixsonne|work, PR is updated, there is work for Travis.16:14
@wikingsonne|work: ping16:15
sonne|workwiking: ?16:16
@wikingsonne|work: so the thing is now that currently when one does make in the src with cmake then everything is going to be compiled16:17
@wikingmeaning unit tests, examples everything16:17
@wikingif of course example building or testing is enabled16:17
sonne|workwiking: could you make parts optional?16:18
@wikinghence if one builds a python modular interface and testing is enabled16:18
@wikingthen unit-test + libshogun examples will be built16:18
@wikingand this is a lot of time16:18
@wikingsee the python modular interface currently is running for 26 min 58 sec16:18
@wikingon travis16:19
@wikingand still not ready16:19
@wikingyeah i mean i can remove parts from make all target16:19
sonne|workyeah I would prefer that this is only done when requested16:19
@wikingwell the request is16:19
sonne|workI mean unit tests examples etc16:19
@wikingor ENABLED_EXAMPLES16:19
@wikingthis you have to give to cmake16:19
@wikingso they are by default OFF16:19
@wikingbut of course this is on on travis16:20
@wikingi was thinking about making some hack variables16:20
@wikingthat would disable parts of this16:20
sonne|worksounds good then16:20
@wikingso we could have16:21
@wikingso that when i'm doing on travis something16:21
@wikingthen we dont rerun unit tests16:21
@wikinglike always16:21
@wikingsame with the libshogun examples16:22
@wikingas currently it is compiled always16:22
@wikingand run ...16:22
@wikingas with ./configure ...16:22
@wikingwhich is of course by default is a good thing16:22
@wikingbut on travis not...16:22
@wikingok i'll add these hack variables16:23
@wikingmaybe i'll add it like16:23
@wikingso that we know that it's really only meant to be used on travis...16:23
sonne|workwiking: maybe you can use the logic that you don't run libshogun examples/unit tests if other interfaces are configured?16:24
sonne|workor at least optionally off16:24
@wikingsonne|work: why ? i mean they should be by default16:24
@wiking*on by default16:24
@wikingif you want examples/tests and building a new shogun16:24
@wikingyou should build all the related examples/tests16:25
@wikinglibshogun + modular stuff you've chosen16:25
@wikinga user would need this16:25
@wikingbut not travis...16:25
sonne|workyes ok so optionall off16:26
@wikingerror: 'modshogun' undefined near line 1 column 116:50
@wikinganybody here who has a working octave_modular env?17:10
@iglesiasgwiking, back some days ago I tried to build octave modular with CMake and it didn't work out17:11
@iglesiasgmaybe you remember about it17:11
@wikingbut do you have a non cmake octave_modular that is working ? :)17:11
@wikingsince i have the for octave17:12
@wikingbut the examples are dying by not finding it17:12
@iglesiasgI can try to build it now if you want17:12
@wikingalthough i set OCTAVE_PATH17:12
@wikingso i dont know anymore wtf is happening17:12
@wikinghow can i get the examples working17:12
gsomixsonne|work, ok, tests works, but Travis says: "No output has been received in the last 10 minutes, this potentially indicates a stalled build or something wrong with the build itself."17:36
gsomixsonne|work, can we merge it? or..?17:36
-!- HeikoS [] has quit [Quit: Leaving.]17:37
-!- lisitsyn [] has quit [Quit: Leaving.]17:47
@iglesiasgall right guys, see you later18:20
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:60ed:c7da:bf1a:7110] has quit [Quit: Ex-Chat]18:21
gsomixwiking, hey18:22
-!- shogun-notifier- [] has quit [Quit: transmission timeout]18:45
gsomixwiking, how can I say to Travis that he should rebuild/test my commit?18:49
gsomixwiking, btw can you merge my commit with fix for CSVFile' test?18:50
@wikinggimme a test that u wanna have rebuilt18:52
@wikingwhich job or build u want on travis?18:53
-!- gsomix_ [~gsomix@] has joined #shogun18:54
-!- gsomix [~gsomix@] has quit [Ping timeout: 260 seconds]18:55
@wikinggsomix: just tellme what u want rebuilt18:55
gsomix_wiking, python_modular + examples19:00
-!- van51 [] has quit [Read error: Connection reset by peer]19:00
@wikingwhich build19:05
-!- gsomix_ [~gsomix@] has quit [Ping timeout: 256 seconds]19:05
@wikinggimme the link for the travis build19:05
-!- van51 [] has joined #shogun19:11
-!- gsomix [~gsomix@] has joined #shogun19:13
-!- shogun-notifier- [] has joined #shogun19:18
shogun-notifier-shogun: Viktor Gal :feature/CMake * ac27f59 / / (6 files):
shogun-notifier-shogun: Fix octave_modular library name19:18
shogun-notifier-shogun: Add options to disable build of unit tests and libshogun examples19:18
shogun-notifier-shogun: Fix target names for csharp examples19:18
-!- lisitsyn [] has joined #shogun19:22
thoralfHey Sergey19:28
shogun-notifier-shogun: Evgeniy Andreev :develop * 752b7d4 / / (5 files):
shogun-notifier-shogun: fix for CSVFile unit-test19:31
shogun-notifier-shogun: Sergey Lisitsyn :develop * 0e037a1 / / (5 files):
shogun-notifier-shogun: Merge pull request #1377 from gsomix/feature/csvfile19:31
shogun-notifier-shogun: Fix for CSVFile' unit-test19:31
lisitsynthoralf: hey19:32
shogun-buildbotbuild #1685 of deb1 - libshogun is complete: Success [build successful]  Build details are at
-!- gsomix [~gsomix@] has quit [Ping timeout: 241 seconds]19:41
shogun-buildbotbuild #1374 of bsd1 - libshogun is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #1175 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Evgeniy Andreev <>19:50
-!- iglesiasg [~iglesias@2001:6b0:1:1041:5137:401d:ae8b:f482] has joined #shogun19:56
-!- mode/#shogun [+o iglesiasg] by ChanServ19:56
-!- gsomix [~gsomix@] has joined #shogun19:59
@wikingwho wrote this..20:03
@wikingit fails bigtime20:03
lisitsynwiking: pickle2720:04
lisitsynwiking: segfault or what?20:04
@wikinglisitsyn: yes20:04
lisitsynwiking: SEGFAULT?20:04
lisitsynokay okay20:04
@wikingbigtime segfult20:05
lisitsynlet me compile and check20:05
@wikinglisitsyn: it worked though on another ubuntu20:05
lisitsynwiking: like you have no machine anymore?20:05
@wikinglisitsyn: travis...20:05
@iglesiasglisitsyn, do you happen to know what is the exact use of IGNORE_IN_CLASSLIST?20:05
lisitsynso travis is dead20:05
lisitsyniglesiasg: YES20:05
@wikingiglesiasg: not to have in class_list.cpp20:05
lisitsynja wiking ist richtig20:05
lisitsynalles grune?20:12
@wikinglisitsyn: well reasonably20:12
-!- iglesiasg [~iglesias@2001:6b0:1:1041:5137:401d:ae8b:f482] has quit [Quit: Ex-Chat]20:15
-!- zxtx [] has quit [Ping timeout: 276 seconds]20:21
-!- van51 [] has left #shogun ["QUIT :Leaving."]20:23
@wikingand now the monkey comes20:24
shogun-notifier-shogun: Viktor Gal :develop * fa25547 / src/ (16 files):
shogun-notifier-shogun: LibShogun with cmake20:25
shogun-notifier-shogun: Viktor Gal :develop * 76de957 / / (32 files):
shogun-notifier-shogun: Move cmake files to the root of the project20:25
shogun-notifier-shogun: add unit test support with cmake20:25
shogun-notifier-shogun: Viktor Gal :develop * 4902996 / .travis.yml,CMakeLists.txt,src/shogun/CMakeLists.txt,tests/unit/CMakeLists.txt:
shogun-notifier-shogun: Fix GMOCK source setting.20:25
shogun-notifier-shogun: change travis config for cmake20:25
shogun-notifier-shogun: Viktor Gal :develop * a6f1f16 / CMakeLists.txt:
shogun-notifier-shogun: Fix gcc version checking20:25
shogun-notifier-shogun: Require 2.8.8 cmake or newer20:25
shogun-notifier-shogun: Viktor Gal :develop * db1e974 / .travis.yml:
shogun-notifier-shogun: Fix travis exclude lines for cmake20:25
shogun-notifier-shogun: Viktor Gal :develop * 5ba076f / CMakeLists.txt,cmake/CheckCompiler.cmake:
shogun-notifier-shogun: Add workaround for compiler version check to support 2.8.7 cmake20:25
shogun-notifier-shogun: Viktor Gal :develop * 54a02e1 / / (4 files):
shogun-notifier-shogun: Fix Eigen3 package name and backport FindJava, FindLibLZMA and20:25
shogun-notifier-shogun: FindDoxygen from 2.8.820:25
shogun-notifier-shogun: Viktor Gal :develop * 97bfa44 / / (4 files):
shogun-notifier-shogun: Fix backporting of package finders from 2.8.820:25
shogun-notifier-shogun: WIP: java20:25
shogun-notifier-shogun: Viktor Gal :develop * e4dbc7b / CMakeLists.txt,cmake/FindSpinlock.cmake,cmake/spinlock-test.cpp:
shogun-notifier-shogun: Merge branch 'feature/CMake' of into feature/CMake20:25
shogun-notifier-shogun: Sergey Lisitsyn :develop * 8e6ace4 / cmake/FindSpinlock.cmake,cmake/spinlock-test.cpp,cmake/spinlock-test-darwin.cpp:
shogun-notifier-shogun: Darwin spinlocks test20:25
shogun-notifier-shogun: Sergey Lisitsyn :develop * f3582a4 / .gitignore:
shogun-notifier-shogun: Updated gitignore to ignore cmake cache20:25
shogun-notifier-shogun: Sergey Lisitsyn :develop * 4649493 / cmake/FindSpinlock.cmake:
shogun-notifier-shogun: Made spinlocks not found if pthreads are not detected20:25
shogun-notifier-shogun: Viktor Gal :develop * ce0e8fb / cmake/spinlock-test-darwin.cpp:
shogun-notifier-shogun: Fix spinlock-test-darwin.cpp for spinlock detection20:25
shogun-notifier-shogun: Viktor Gal :develop * 54232cc / / (17 files):
shogun-notifier-shogun: add lua, csharp, java, octave modular support20:25
shogun-notifier-shogun: Viktor Gal :develop * 0f30d1f / CMakeLists.txt:
shogun-notifier-shogun: Restructure interface choosing in cmake20:25
shogun-notifier-shogun: Viktor Gal :develop * a152427 / .travis.yml,CMakeLists.txt:
shogun-notifier-shogun: add missing endif in cmake20:25
shogun-notifier-shogun: add preliminary support for cpack20:25
lisitsynwiking: so this is happening20:25
@wikinghave fun20:25
shogun-notifier-shogun: Fix version date macros in versionstring.h20:25
shogun-notifier-shogun: fix ld_library_path env setting in python_modular example20:25
shogun-notifier-shogun: Viktor Gal :develop * c2a6966 / examples/undocumented/libshogun/CMakeLists.txt:
shogun-notifier-shogun: fix libshogun examples' working directory20:25
shogun-notifier-shogun: Viktor Gal :develop * 2660db7 / tests/unit/CMakeLists.txt:
shogun-notifier-shogun: Build unit tests with 'make' if unit testing is enabled20:25
shogun-notifier-shogun: Viktor Gal :develop * 5ab4f9c / .travis.yml,CMakeLists.txt:
shogun-notifier-shogun: Add ENABLE_TESTING option to cmake20:25
shogun-notifier-shogun: ctest should be enabled with ENABLE_TESTING regardless20:25
shogun-notifier-shogun: whether GMOCK or GTEST framework has been found20:25
shogun-notifier-shogun: Viktor Gal :develop * 9da75b5 / .travis.yml:
-!- shogun-notifier- was kicked from #shogun by wiking [shogun-notifier-]20:25
shogun-notifier-shogun: Use ctest instead of 'make test'20:25
shogun-notifier-shogun: This way we can provide error output of failed test with the --output-on-failure command line arg20:25
@wikingout with u20:26
@wikingthat's just too much of a commit20:26
@wikingcmake is merged20:26
lisitsynwiking: that was THE EFFORT20:26
-!- gsomix [~gsomix@] has quit [Ping timeout: 246 seconds]20:32
shogun-buildbotbuild #1479 of deb3 - modular_interfaces is complete: Failure [failed test ruby_modular]  Build details are at  blamelist: Evgeniy Andreev <>20:32
-!- travis-ci [] has joined #shogun20:35
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []20:35
shogun-buildbotbuild #1176 of cyg1 - libshogun is complete: Failure [failed compile]  Build details are at  blamelist: Sergey Lisitsyn <>20:38
@wikingshort howto sent to the mailing list20:41
@wikingfuck all20:41
@wikinglisitsyn: can u please rewrite the README file based on that howto? :D20:45
lisitsynwiking: I guess INSTALL20:46
-!- gsomix [~gsomix@] has joined #shogun20:46
@wikingmmm well the readme is not the best as well20:46
@wikingi mean either20:46
@wikingbut yeah20:46
@wikingsomebody should write a nice output20:47
@wikingof the cmake20:47
@wikingwith MESSAGE()20:47
@wikingwhat r the libraries detected (flags etc)20:48
lisitsynwiking: could you add issues for that?20:48
@wikingwhich interfaces are turned on etc20:48
@wikinghehe yeah i'll close the general cmake issue now20:48
-!- travis-ci [] has joined #shogun20:49
travis-ci[travis-ci] it's Sergey Lisitsyn's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []20:49 integration test errors20:58
@wikingwonder why is this:21:00
@wiking-[  9.92749338e-02   7.44443696e-01   8.76619850e-06]21:00
@wiking+[  9.92749338e-02   8.25347020e-01   8.76619850e-06]21:00
@wikingalmost ;)21:00
lisitsynwiking: ahhh slight miss21:01
@wikingstill it's a miss :P21:01
@wikingheheh but things are getting greeen21:01
lisitsynwiking: whole kaliningrad area has blackout :D21:02
@wikingthat's a big segfault :)21:02
@wikingheheh one can create an eclipse project with cmake :P21:05
shogun-buildbotbuild #1687 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Sergey Lisitsyn <>, Viktor Gal <>21:17
shogun-buildbotbuild #1480 of deb3 - modular_interfaces is complete: Success [build successful]  Build details are at
shogun-notifier-shogun: Viktor Gal :develop * 922d28c / tests/unit/mathematics/
shogun-notifier-shogun: Fix Complex::complex_test precision.21:21
shogun-notifier-shogun: The test failed on OSX due to precision error. changing from 1E-14 to21:21
shogun-notifier-shogun: 1E-1321:21
shogun-buildbotbuild #1688 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Viktor Gal <>21:29
-!- travis-ci [] has joined #shogun21:31
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []21:31
@wikingok cpacking21:48
@wikingsoon meeting?21:48
lisitsynwiking: should be21:48
-!- HeikoS [~heiko@] has joined #shogun22:00
-!- mode/#shogun [+o HeikoS] by ChanServ22:00
lisitsynHeikoS: hey there22:01
@HeikoSlisitsyn: hi!22:02
-!- iglesiasg [] has joined #shogun22:02
-!- mode/#shogun [+o iglesiasg] by ChanServ22:02
@iglesiasghello guys22:02
@wikingjust a sec22:02
@iglesiasgsorry for the delay22:02
@iglesiasgdid I miss something?22:02
@wikingi need a coffee+cigarette22:02
@wikingbut u can start discussing22:02
@wikingabout the AWESOMENESS OF CMAKE!22:02
@HeikoSso all there?22:02
@HeikoSiglesiasg, wiking, lisitsyn22:02
@wikingand if somebody could fix the integration test that'd be great22:02
@iglesiasgsonney2k won't be here with us finally?22:03
@HeikoSiglesiasg: no he cannot be here22:03
@HeikoSSo first of all a few words about these meetings:22:03
lisitsynI am here22:03
@HeikoSs?ren and I had the idea it would be nice to have regular discussion like the one a while ago22:03
@HeikoSand document the outcomes22:03
@HeikoSbecasue during the day there are always so many other things22:03
@HeikoSso I suggest we take turns in organising them every month or so22:04
@HeikoSSo I got the following points for today22:04
@HeikoS1.) Talk a bit about how we organise the meetings (documentation etc)22:04
@HeikoS2.) Talk about docsprint application22:04
@HeikoS3.) Talk about wiking's amazing cmake job22:04
@HeikoSany other points you want to talk about?22:05
lisitsynI have nothing else22:05
@wiking+1 foundation22:05
@HeikoSwiking: ok good idea, lets do that as 4.)22:05
@HeikoSso 1.)22:05
@iglesiasg(I am fine with those too)22:05
@HeikoSI will write a short summary of the things decided today and put them in the github wiki22:05
@HeikoSso that we have a growing list of ideas22:06
@HeikoSwho will do the next one?22:06
@HeikoSwiking: you?22:06
@HeikoSnice :)22:06
@HeikoSso then docsprint22:06
@HeikoSI drafted this little application, deadline is tomorrow22:07
@HeikoSLast year we did not get in, I think one reason might have been that our proposal was too mathy22:07
@HeikoStoo much ML22:07
@HeikoSThis year, this is only one of a few points22:07
@HeikoSSo please edit the things and write comments22:07
@HeikoSmake your own description more fancy, shorten my scentences (I tend to write too much)22:08
@HeikoSso that we have a nice proposal. I will submit this tomorrow noon22:08
@HeikoSAny comments on the thing already?22:08
@iglesiasgnice work, I didn't get time yet to read it completely though22:08
@iglesiasgwill do after the meeting22:08
@wikingi haven't managed to read it yet sorry :(22:08
@HeikoSplease read carefully and tune it22:09
@iglesiasgjust glanced over it22:09
@HeikoSalwasys better maybe you have some more ideas22:09
@HeikoSalso have a look at last years docsprint22:09
@HeikoSthere are many blog entries about it and one can also read the book e.g. by KDE22:09
-!- zxtx [] has joined #shogun22:09
@HeikoSI suggested 4 points to document22:10
@HeikoSinstallation - framework overview - a handful of ML algos - hacking shogun(dev guidelines)22:10
@HeikoSKDE last year have like a little webpage with the book22:10
@HeikoSI think for part 1,2,4, we should have a simmilar thing, maybe orient on scikit learn's page22:11
@HeikoSfor 3 we should write ipython notebook and then include the rendered version in the book22:11
@HeikoSwiking, lisitsyn, iglesiasg so please add your stuff and make the proposal awesome, you will get a free trip to US if this works and we can spend three nice days on this22:12
lisitsynokay I'll take a look for sure22:12
@HeikoSalso look at this one22:12
@HeikoSok then, wiking tell us about cmake :)22:13
@wikingok so it's merged22:13
@wikingi've completely refactored travis22:13
@wikingand as u can see now we have still 2 jobs failing22:13
@wikinga) python integration with fisher shit22:13
@wikingb) one octave example22:14
@wikingoctave modular wasn't in previously in travis so maybe22:14
@wikinguntil it doesn't get fixed we should move it to allowed to fail jobs22:14
@wikingsince as u can see we have now jobs that are allowed to fail22:14
@wikinglike the new OSX build22:14
@wikinganyhow i wrote a short tutorial on the mailing list how to use cmake for building22:15
@HeikoSwiking: could you maybe add a little README on how things work?`or extend the existing one22:15
@wikingwe need still some lines about this in the README/INSTALL files22:15
@wikingcan somebody please do this?22:15
@wikingbased on that little tutorial22:15
@wikingi'm like fuckk off all with this cmake now22:15
@HeikoSwiking: just running cmake for the first time :)22:16
@HeikoSwiking: haha :)22:16
@wikingthere are several things we need to discuss22:16
@HeikoSwiking: why do I have to create a dir for it?22:16
@HeikoSha! it works! :)22:16
@wiking1) changing buildbots to use cmake. where we not only want to compile but with the nightly bots22:16
@wikingwe want to create packages with cpack and upload them somewhere22:16
@wikingthese packages would be: tar.gz, tar.bz2, .deb and .rpm22:17
@HeikoSwiking: does it create the deb etc already?22:17
@wikingthe first to is pretty easy to create22:17
@wikingmake package22:17
@wikingthe .deb and .rpm creating still needs hacking22:17
@wikingso basically the CMakeLists.txt still needs some love to do those properly22:17
@wikingbut i could already create libshogun.deb and libshogun-dev.deb22:18
@wikingwhich is something22:18
@HeikoSwiking: cool yes22:18
@HeikoSwiking: its fast22:18
@wikingwe need to set up a repository for this22:18
@HeikoSwiking: how can I do --enable-debug and disable optimization22:18
@HeikoSalso do I need to specify gtest?22:18
@wikingto cmake22:18
@wikingdebug is by default afaik... but let me know if not :)22:19
@wikingif you do -DENABLE_TESTING=ON22:19
@iglesiasgHeikoS: have a look to ccmake too22:19
@wikingthen it'll automagically download the right version of gmock/gtest22:19
@wikingso you don't have to care about that at all22:19
@HeikoSwiking: how to enable debug=22:19
@wikingHeikoS: debug is by default22:19
@HeikoSwiking: ok, so how can I print a list of all options like for configure?22:19
@wikingHeikoS: cmake by default supports 2 build types: Debug and Release22:19
@wikingHeikoS: ccmake ..22:20
@wikingor cmake gui22:20
lisitsyncmake -i22:20
@wikingthat'll show all the options u want/need22:20
lisitsynworks too22:20
@wikingah yeah22:20
@wikingif ccache is found22:20
@HeikoSwiking: I have a problem:22:20
@wikingit's automatically enabled22:20
@wikingHeikoS: tell me22:21
@HeikoScmake -DPythonModular=ON  -DDISABLE_OPTIMIZATIONS=ON ..22:21
@HeikoSCMake Error: Attempt to add a custom rule to output "/home/heiko/Desktop/shogun/shogun/build/src/interfaces/python_modular/Converter_includes.i.rule" which already has a custom rule.22:21
@HeikoSand many more22:21
@wikingHeikoS: yes22:21
@HeikoS-- Could NOT find Eigen3: Found unsuitable version "3.0.93", but required is at least "3.1.2" (found /usr/include/eigen3)22:21
@wikingHeikoS: that shit is because of the remaining files of ./configure22:21
@HeikoSdoesnt it download eigen? How can I add that?22:21
@HeikoSah I see so make clean22:21
@wikingwhat's the command to reset the repo and remove all the files that are not part of the repo?22:22
@wikingor yeah22:22
@wikingin src do a make clean22:22
@HeikoSwiking: worked22:22
@HeikoSand the eigen thing?22:22
@wikingHeikoS: as written in the email22:22
@wikingHeikoS: if u dont have the proper eigen version22:22
@wikingthen just -DBUNDLE_EIGNE=ON22:22
@wikingthen just -DBUNDLE_EIGEN=ON22:22
@HeikoSah cool22:22
@HeikoSwiking: cant this be done autonmatically?22:22
@wikingHeikoS: what?22:23
@wikingHeikoS: you mean if there's no right version of eigen?22:23
@HeikoSwiking: like if eigen3 is not there this is added automagic?22:23
lisitsyngit clean -dfx22:23
lisitsynis the command22:23
@wikingwell i dont know... maybe somebody wants shogun w/o eigen22:23
@HeikoSwiking: mmmh default should be with22:23
@HeikoSand then turn off if desired22:24
@wikingHeikoS: that can be done22:24
@wikingbut currently we have a problem22:24
@wikingwith cmake22:24
@HeikoSah cool its compiles eigen :)22:24
@wikingsay you add a new class22:24
lisitsynit won't be visible22:24
lisitsynbecause of cmake stupidity22:25
@wikingimho there's a way around this22:25
@wikingusing some hacks22:25
@wikingbut i hadn't had the time to fix this22:25
@wikingso if u add a new file22:25
@iglesiasgwhat is the way to do then?22:25
@iglesiasgmake clean?22:25
@wikingu have to re-run cmake22:25
@iglesiasgall righ22:25
@HeikoSwiking: thats not too bad22:26
@wikingyeah maybe22:26
@wikingi'm suspicious about the versionheader.h22:26
@wikingas well22:26
@wikingbecause it might trigger the rebuild of the whole shit... but i dont know22:26
@wikinghave to test it22:26
@wikinglet me know if you are having too much rebuilds :P22:27
@wikingand i'll try to fix it asap22:27
@HeikoSwiking: cool22:27
@wikinganyhow it's there and it's more or less working22:27
@HeikoSwiking: awesome job you did there :)22:27
@wikingstatic interfaces are completely missing22:27
@wikingif anybody has the feeling for doing it go ahead22:27
@HeikoSwiking: maybe put a TODO list in a github issue so that we know whats missing etc22:27
@wikingthat should be really the easier part afaik22:27
@wikingwill do22:27
@HeikoSwiking: cool!22:28
@HeikoSany more on cmake then?22:28
@wikingmmm yeah22:28
@wikingit's not that much related22:28
@wikinglets fucking fix this error in integration22:28
@wikingnot tomorrow22:28
@HeikoSvolunteers? :)22:28
@wikingso that we can have finally green buttons on PR22:28
@iglesiasgwiking: this one?
@wikingi can put the octave modular interface into allowed to fail22:29
@wikingand then it should be good22:29
@HeikoSwiking: how to best access this remote report thing? the travis ui is always so slow22:29
@wikingHeikoS: well i've created this:22:29
@HeikoSresults actually changed in the test22:29
@HeikoS7.4 vs 8.2522:30
@HeikoSso just bookmark the site?22:30
@wikingi still have to work on this22:30
@HeikoSto travis for now22:30
@wikingto get all the configure shit and stuff there22:30
@wikingbut as soon as it's ready22:30
@HeikoSsince when is the integration test failing?22:30
@wikingi'll send a message about this22:30
@HeikoSlisitsyn: would you mind fixing the test?22:31
@wikingyes this22:31
@HeikoSI have to fix another one  :)22:31
@wiking-[  9.92749338e-02   7.44443696e-01   8.76619850e-06]22:31
@wiking+[  9.92749338e-02   8.25347020e-01   8.76619850e-06]22:31
lisitsynI don't mind22:31
lisitsynwhat test?22:31
@HeikoSwiking: just pasted it22:31 changed results22:32
@wikinglisitsyn: The following tests failed! setting 1/122:32
@wikingDetailled failures: setting 1/122:32
@HeikoSquite a bit22:32
@HeikoSso why could that be?22:32
lisitsynno idea what it is22:32
@wikingi'm just running now the same thing on a different ubuntu22:32
@HeikoSanyone HAS an idea? (I dont?)22:32
@wikinglet's see what it does22:32
lisitsynthat's something sonney2k added in september of 2011 IIRC22:32
@HeikoSwiking: so random generator maybe?22:32
@HeikoSso  he should fix it ....22:32
@wikingHeikoS: dunno... i dont think fisher has anything to do with prng22:33
@HeikoSbut data maybe?22:33
@HeikoSsince when does it fail?22:33
@wikingit's a numpy binary22:33
@wikingHeikoS: lemme check travis22:33
@wikingbefore the merge22:33
@HeikoSwiking, lisitsyn on lets fix after meeting22:34
@wikingHeikoS: btw travis is much more responsive with chrome22:34
@HeikoSlast point: foundation22:34
@HeikoSso we collected all the signatures on the current state of the Satzung22:34
@wikingbtW: before the merge it was not failing!22:34
@HeikoSThere were a few updates to it, so I will update this22:34
@HeikoSI am away next week, but once I am back, I can ask sonney2kto send me the copy of the things etc22:35
@HeikoSThen I will check the next steps22:35
@HeikoSWe have to write a protocol about how the foundation was funded22:35
@HeikoSah one thing22:35
@HeikoSwiking, lisitsyn, iglesiasg do you got a good idea how to do an anyonymous vote online?22:36
@wikingHeikoS: how official does it have to be?22:36
@HeikoSBecause we have to vote for a main guy for the foundation22:36
@HeikoSnot official22:36
@iglesiasgwe can all get into a chat22:36
@iglesiasgwith guest_ names or so22:36
@wikingand it has to b anonymous?22:36
@HeikoSjust that we can write in the protocol that there was an anonymous vote22:36
@wikingiglesiasg: good one ;)22:36
@HeikoSbut everyon just allowed to vote once22:37
@wikingyeah random irc channel22:37
@HeikoSno then we cannot controll the votes22:37
@iglesiasgIRC through webchat in that case22:37
@HeikoSand it would be good if we can do it over a few days22:37
@HeikoSno we have to invite people22:37
@iglesiasgwell just a doodle then22:37
@HeikoSand then they are allowed to vote once22:37
@iglesiasgaaah wait only once22:38
@wikingHeikoS: it's open to everybody?22:38
@HeikoSwiking: no invite only22:38
@wikingHeikoS: ah ok22:38
@HeikoSlisitsyn: its not legal otherwise22:38
@wikingHeikoS: then i think it's good22:38
lisitsynyeah I understand HeikoS22:38
@wikingHeikoS: random irc channel guest22:38
@wikingand then n number of ppl22:38
@wikingand we cound the number of votes22:38
@wikingthey should match22:38
@HeikoSwiking: but this channel needs to run for a dew days22:38
@HeikoSwiking: if we send the link to the people it doesnt matter how many vote22:39
@HeikoSthats good22:39
@HeikoSbut it needs to be logged22:39
@HeikoSand its non anon22:39
@HeikoSIPs are visible22:39
@HeikoSiglesiasg: could you check that out and implement one for us. Just make a little test22:39
@wikingHeikoS: ^22:39
@HeikoSok once this is working, you all vote for or against me and then we can collect the protocol and all the paperwork and sonney2kcan bring it to the court in Berlin22:40
@HeikoSwe have to pay some money22:40
@HeikoSthen we can apply to be recognised as charity22:40
@sonney2kwiking, please put that description of cmake you did in the readme on how to compile shogun22:40
@HeikoSand then we have  a legal body that is allowed to take money without paying taxes22:40
@HeikoSthats it22:41
@wikingsonney2k: i've propagated the job to somebody else :)22:41
@HeikoSso next week22:41
lisitsynyeah me me22:41
@sonney2kwiking, lisitsyn  excellent22:41
@HeikoSok anything else for the foundation then?22:41
@wikingsonney2k: buildbots?22:41
@sonney2kHeikoS, yeah we need notar for the copy22:41
@wikingsonney2k: do u have time for that today?22:41
@sonney2kwiking, we can change libshogun based buildbots22:41
@HeikoSsonney2k: yeah, but first vote leader and update Satzung22:41
@wikingsonney2k: u can change modular as well...22:41
@sonney2kwiking, I am rather tired though22:41
@sonney2kHeikoS, you have my vote22:42
@wikingsonney2k: gimme access and i'll make it for u22:42
@HeikoShaha :)22:42
@HeikoSsonney2k, lisitsyn, wiking, iglesiasg anything else for the meeting, comments?22:42
@HeikoSIf not, wiking will do the next one, I suggest in one month22:42
@HeikoSI will put up a summary22:42
@HeikoSmost important: please everyone read and tune the docsprint proposal!22:42
@HeikoS*meeting over22:43
@HeikoSWill do some cooking now, be back in a little while22:43
@sonney2kwiking, then we only need static22:43
@wikingsonney2k: yeps22:43
@wikinglisitsyn: this fucking integration fails big time on another ubuntu :D22:43
@wikingjust generating the log22:44
@iglesiasglisitsyn, HeikoS, wiking : tell me if you got the mail guys22:44
lisitsyniglesiasg: I did22:44
@wikingiglesiasg: gotit22:44
@iglesiasgI think that is pretty much what we need22:44
@HeikoSiglesiasg: got it :)22:45
@sonney2kgsomix, are the tests finally fixed?22:45
@iglesiasgHeikoS: so that does everything we need, right?22:45
@iglesiasgwe can set a time interval, anonymous, and ensure each one votes once22:46
@wikingsonney2k: on the nightly we would need ipython 1.022:47
@wikingsonney2k: can u git clone it?22:47
gsomixsonney2k, yep!22:47
@iglesiasgI need to grab dinner now, I am starving!22:47
@iglesiasgHeikoS: is there anything else to discuss?22:47
@sonney2kgsomix, PR please22:47
gsomixsonney2k, already merged22:48
@wikingwhat the hack is this:22:49
@wikingwhy n and why o? )22:49
@sonney2kargh got disconnected22:52
@sonney2kwiking, did you already change travis?22:52
@wikingsonney2k: yes yes22:52
gsomixsonney2k, already merged,
@wikingthe osx builder machine is quite slow :S22:53
@sonney2kwiking, sth is still weird - the is not giving the same result22:54
@wikingsonney2k: i know...22:54
@wikingbut i dont know why22:54
@sonney2kit uses floatmax_t, lgammal22:56
@sonney2kwiking, do we have a test for lgammal ?22:57
@sonney2kwiking, it has a ifdef HAVE_LGAMMAL22:57
gsomixsonney2k, changes and new code will be tomorrow then. need sleep a little.22:57
gsomixnite guys22:57
@sonney2kgsomix, ok cu tomorrow!22:58
@wikingsonney2k: -- Looking for lgammal22:59
@wiking-- Looking for lgammal - not found22:59
@wikingsonney2k: Checking for Math lgammal() support ... yes23:01
@wikingok checking the error why it doesn't find it23:01
-!- travis-ci [] has joined #shogun23:02
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []23:02
-!- HeikoS [~heiko@] has quit [Quit: Leaving.]23:05
-!- gsomix [~gsomix@] has quit [Ping timeout: 264 seconds]23:07
shogun-notifier-shogun: Soeren Sonnenburg :develop * 4daf5ee / examples/undocumented/python_modular/ (4 files):
shogun-notifier-shogun: convert more examples to use CSVFile23:09
@sonney2kargh got dc again23:10
@sonney2kconnection in train is not that stable today :/23:11
@sonney2kwiking we have a Makefile in shogun root23:11
@sonney2kfor the release23:11
@sonney2kwill that be overwritten?23:11
@wikingmkdir build23:12
@wikingcd build23:12
@wikingcmake ../23:12
@sonney2kyeah just seen it23:12
@sonney2kgot to leave train now23:12
-!- gsomix [~gsomix@] has joined #shogun23:19
shogun-notifier-shogun: Viktor Gal :develop * 2c8a285 / CMakeLists.txt:
shogun-notifier-shogun: Require libm for compilation on UNIX hosts23:24
@wikingok lets see what this does23:24
-!- gsomix [~gsomix@] has quit [Ping timeout: 256 seconds]23:26
-!- gsomix [~gsomix@] has joined #shogun23:28
shogun-buildbotbuild #1690 of deb1 - libshogun is complete: Failure [failed test]  Build details are at  blamelist: Viktor Gal <>23:28
-!- gsomix [~gsomix@] has quit [Ping timeout: 264 seconds]23:35
@wikingsonney2k: ping23:37
@sonney2kwiking, pong?23:39
@wikingsonney2k: can we setup a nightly on fbsd?23:40
@wikingso then i can start hacking with the ipython notebooks23:40
-!- zxtx [] has quit [Ping timeout: 246 seconds]23:43
@sonney2kwiking, ahh no need for that I will do that on the nightly buildbot so we can directly have it on the page23:47
@wikingi'll setup the new target23:48
@sonney2kwiking, we need some way of listing the notebooks / and decision where we put it on the website23:48
@sonney2kwiking, ohh not a cmake target though23:48
@wikingyou'll only have to give the path to the ipython dir where u cloned the latest ipython23:48
@sonney2kwiking, I wanted to run the heiko gave me23:48
-!- HeikoS [~heiko@] has joined #shogun23:49
-!- mode/#shogun [+o HeikoS] by ChanServ23:49
@wikingsonney2k: yeah u need ipython 1.0 for that23:49
@sonney2kwiking, and I already 'fixed' the script to fail when there is an exception in the notebook23:50
@sonney2kcurrently it will never list failures23:50
@sonney2kwiking, biggest issue currently is imho where to put it on the website and to do it with some style23:50
@sonney2kwiking, like here
@sonney2kimporterror but no failures23:51
@HeikoSsonney2k: does running the notebooks and uploading work already?23:53
@HeikoSsonney2k:  what do you think on the idea to include some notebooks in the docsprint? for one of the parts?23:53
@sonney2kHeikoS, I mean format wise we should use the notebooks - iglesiasg showed me some book fully done in notebook23:54
@HeikoSsonney2k: yeah I know23:55
@sonney2kwiking, any idea about the lgammal?23:55
@HeikoSah cool23:55
@iglesiasgHeikoS: did you see it?23:55
@HeikoSsonney2k: there are some things however for that we need other listings than python23:55
@iglesiasgHeikoS: I can dig in for the link otherwise23:55
@HeikoSso I would only do the ML framework related stuff in there23:55
@HeikoSnot the install shogun and the hacking shogun23:55
@wikingsonney2k: it's fixed with the last commit23:55
@HeikoSonly framework/class overview and algo description23:56
@wikingsonney2k: and it actually fixed that error in the integration test23:56
@HeikoSiglesiasg: see what?23:56
@wikingsonney2k: but you've created 2 new errors23:56
@iglesiasgHeikoS: the complete book in ipython23:56
@wikingsonney2k:                                ERROR (name 'LoadMatrix' is not defined)23:56
@sonney2kwiking, how do I force cmake to do a full reconfigure23:58
@wikingrm -rf  *23:58
@wikingin the build23:58
@sonney2kno other way?23:59
@wikingwell that's one way23:59
@sonney2kwiking, yeah but a typo is dangerous23:59
@sonney2kso I would rather want to do that with some cmake option23:59
--- Log closed Fri Aug 09 00:00:01 2013