Open in new window / Try shogun cloud
--- Log opened Fri Aug 05 00:00:52 2016
-!- OXPHOS [4441a916@gateway/web/freenode/ip.] has joined #shogun01:43
@wikingOXPHOS: y002:30
@wikinglemme look in the rs02:30
@wikingi should be able to run a full buildbot check on that feature branch02:30
@wikingas i'm afraid we might not catch some errors :D02:31
-!- shogun-notifier- [] has joined #shogun02:33
shogun-notifier-shogun: OXPHOS :feature/linalg_refactor * 05ee97d / src/shogun/mathematics/linalg/ (3 files):
shogun-notifier-shogun: add return value/fix msgs/remove white lines02:33
shogun-notifier-shogun: Viktor Gal :feature/linalg_refactor * 337d2f3 / src/shogun/mathematics/linalg/ (3 files):
shogun-notifier-shogun: Merge pull request #3404 from OXPHOS/linalg_refactor_202:33
shogun-notifier-shogun: LinalgRefactor - minor fix - warnings02:33
shogun-notifier-shogun: OXPHOS :feature/linalg_refactor * 00fd311 / / (5 files):
shogun-notifier-shogun: linalg-matrix block sum02:34
shogun-notifier-shogun: Viktor Gal :feature/linalg_refactor * 15a5ae9 / / (5 files):
shogun-notifier-shogun: Merge pull request #3403 from OXPHOS/linalg_sum_block02:34
shogun-notifier-shogun: LinalgRefactor - matrix block sum02:34
shogun-notifier-shogun: OXPHOS :feature/linalg_refactor * 82989a6 / / (6 files):
shogun-notifier-shogun: linalg-sgvector-add-inplace02:35
shogun-notifier-shogun: Viktor Gal :feature/linalg_refactor * acaeca4 / / (6 files):
shogun-notifier-shogun: Merge pull request #3391 from OXPHOS/linalg_add_inplace02:35
shogun-notifier-shogun: LinalgRefactor -  SGVector - inplace add02:35
OXPHOSwiking: thx! I'll fix the other linalg prs soon.03:23
-!- travis-ci [] has joined #shogun03:29
travis-ciit'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 []03:29
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: leaving]03:39
-!- travis-ci [] has joined #shogun04:05
travis-ciit'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 []04:05
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun04:09
-!- mode/#shogun [+o wiking] by ChanServ04:09
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: leaving]04:13
-!- wiking [] has joined #shogun04:20
-!- wiking [] has quit [Quit: ZNC 1.6.3 -]04:24
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun04:25
-!- mode/#shogun [+o wiking] by ChanServ04:25
@wikingshogun-buildbot: dance04:35
-!- travis-ci [] has joined #shogun04:52
travis-ciit'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 []04:52
-!- shogun-notifier- [] has quit [Quit: transmission timeout]05:35
-!- OXPHOS [4441a916@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]05:39
-!- sanuj [~sanuj@] has joined #shogun07:40
sanujwiking, i have updated this
@wikingcan you please explain more detail about07:41
@wikingwhat's TagRealVector and RealVector are like? and why not the same? :)07:41
sanujTagX contains a name and type information for the object X07:41
sanujso TagRealVector is an identifier for RealVector object07:42
@wikingstill dont understand07:43
@wikingcould you maybe explain it a bit differently?07:43
sanujlet me try07:44
sanujyou have an object of RealVector called r_vec07:44
sanujand you have another object called "obj"07:44
sanujobj has a member_variable named "numbers" which is of type RealVector07:45
sanujnow there are two ways to set the value of member_variable named "numbers" in "obj"07:45
sanujobj.setsRealVector("numbers", r_vec)07:46
@wikingwhy not: obj.sets("numbers", r_vec)07:46
sanujbecause sets<T> is a template function07:47
sanujso in swig you need to instantiate it with different types07:47
@wikingsets is not a sets(string, Any) ?07:47
sanujtemplate <typename T, typename U=void>07:48
sanujvoid sets(const std::string& name, const T& value)07:48
sanujwiking, well, how else can it be?07:49
@wikingsets could accept any type07:49
@wikinganyhow yes listeing07:49
@wikingso you have07:49
@wikingso you07:49
sanujyou have name of parameter that you want to set, value of parameter and type of parameter as T07:50
@wikingsets[WHATEVER THE FUCK YOU WANNA SET](string, obj)07:50
@wikingfor the swig interface?07:50
@wikingso if i want to set for an object07:50
@wikingmy new KurvaMachine07:50
@wikingthen i'll have07:50
@wikinga setsKurvaMachine("name of kurva machine instance", machine)07:50
@wikingthis is the swig interface/07:51
sanujwiking, no, only those params can be set which are registered in the object by using register_param()07:51
sanujin the constructor07:51
@wikingbut say07:51
@wikingi have a07:51
@wikingin my object07:51
@wikingso say my new class is liek07:52
@wikingclass X{ .... CKurvaMachine m_kurva_machine;};07:52
@wikingi mean this is the old way07:52
@wikingbut just to give you an example07:52
@wikingso in this case07:52
@wikingk = X()07:52
@wikingk.setsKurvaMachine("name", kurva_obj)07:52
@wikingi dont know if i wanna laugh or cry07:53
@wikingthis is fucking madness07:53
@wikingbut ok07:53
sanujwiking, you can use sets() also07:53
sanujbut you need to make TagKurvaMachine("name")07:53
sanujk.sets(tag_kurva_machine, kurva_obj)07:54
@wikingso then07:54
@wikingi'll have to have a new Tagged instance07:54
@wikingthat's nothing else07:54
@wikingjust a wrapper07:54
@wiking*ish thingy07:54
sanujwiking, i had discussed this with lisitsyn and HeikoS both07:55
@wikingsanuj, ok08:11
@wikingall green08:11
@wikinglet's merge it then08:11
-!- shogun-notifier- [] has joined #shogun08:11
shogun-notifier-shogun: Sanuj :feature/tags * d15f5b0 / / (12 files):
shogun-notifier-shogun: add swig interface for tags08:11
shogun-notifier-shogun: Viktor Gal :feature/tags * b5b3cbf / / (12 files):
shogun-notifier-shogun: Merge pull request #3357 from sanuj/tags-swig-jinja08:11
shogun-notifier-shogun: add swig for tags using Jinja08:11
sanujwiking, ok so i tried k.sets("name", kurva_obj) on my local08:18
sanujand it works fine08:18
sanujbut the problems is then there won't be symmetry in sets(), gets() and has()08:19
sanujk.sets("name", kurva_obj) will work08:19
sanujbut k.gets("name") will not work08:20
sanujk.getsKurvaMachine("name") and k.gets(tag_kurva_machine) will work08:20
-!- sanuj [~sanuj@] has quit [Ping timeout: 240 seconds]11:02
-!- shogun-notifier- [] has quit [Quit: transmission timeout]11:11
-!- sanuj [~sanuj@] has joined #shogun11:26
-!- sanuj [~sanuj@] has quit [Ping timeout: 258 seconds]11:35
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/] has joined #shogun11:53
@wikingc4goldsw, yoyoyo11:53
@wikingi've merged your lars11:53
c4goldswEarth to wiking, looks like the build worked!11:53
@wikingyep yep11:53
@wikingand it's memleak free \o/11:53
c4goldswThat's what I was looking for11:53
c4goldswwiking doing well?11:54
@wikingyep yep thnx11:54
-!- sanuj [~sanuj@] has joined #shogun12:15
@wikingc4goldsw, how's your refactor going?13:19
c4goldswwiking: Slowly, taking my time.  I did LDA a few days ago but, when doing a unit test, the outputs are completely different for one test and the same for another.13:20
c4goldswIt's odd, but I've yet to put it through a debugger.13:20
@wikingyeah gdb/lldb is usually something that helps a lot13:21
@wikinga good friday read btw:
c4goldswHaha, the title.13:22
c4goldswwiking: school starts in four weeks or so - during that time, I'll try doing at least a class a week.13:22
@wikinggreat man!13:23
@wikinghighly appreciated :)13:23
@wikingwhat's your major in?13:23
c4goldswCS, I want to go into ML13:23
c4goldswI'm going into my third year of undergrad.13:23
@wikingwhich school?13:23
c4goldswU of T, the school hinton works at.13:23
c4goldswGood for ML / AI.13:23
c4goldsw* Hinton13:23
@wikingbtw you know we have a track record of having people getting accepted to MIT :)13:23
@wikingso people contributing (gsoc or otherwise)13:24
c4goldswHahaha, I think the top tier schools are a little out of my reach.  I have a 3.65 right now, hoping to bump it up.13:24
c4goldswI'm aiming for Oxford, Cambridge, Imperial, Edinburgh, etc.13:25
@wikingheheh i have no idea about gpes and sat-s13:25
@wikingjust heard them in movie13:25
@wikingi have a friend working at Edinburgh uni13:25
@wikinghad one in oxford but now he is back in mexico13:26
c4goldswWhere did you study?13:26
@wikingbsc+msc at budapest university of technology13:26
@wikingthen my phd at ghent university, belgium13:26
@wikingand i was for a while a researcher at australian national uni13:26
@wikingnothing fancy :)13:26
c4goldswHow'd you wind up in Singapore?13:27
@wikingmy wife got here a position13:27
@wikingso i came with the luggage13:27
c4goldswSprichst du Deutsch?13:28
@wikingy0 bissl13:28
c4goldswArbietest du als Hausmann jetzt? ;)13:28
@wikinghahahah ja13:28
c4goldswInformatik is mein Hauptfach, Deutsch ist mein Nebenfach.13:29
@wikingwas away13:32
@wikingja, ich habe in wien drei jahre gewohnt...13:32
@wikingso yeah13:32
@wikingbut that was a long time ago13:32
@wikingalmost lost all my german13:32
@wikingmainly i understand still everything13:32
@wikingbut speaking.... especially my writing is horrendeous13:33
sanujwiking, who went to MIT?13:38
@wikingsanuj, i think some of them are still there13:40
sanujthe mocha guy13:41
@wikingi think we have like 2 more13:42
sanujc4goldsw, mine is even worse13:42
sanujlisitsyn, wazzup13:44
@wikingjacob walker is at cmu now13:44
@wikingso it used to be a good thing :)13:45
sanujwiking, yeah13:45
lisitsynsamara state aerospace university here13:45
@wikinglisitsyn, :DDDDDDDDDDDDDDD13:45
@wikingbest of the best13:45
lisitsyn6921 in the world rating13:45
@wikingthose ratings are actually bullshit13:45
lisitsynthe best people13:45
sanujlisitsyn, why did you do masters?13:46
@wikingSaurabh7, ping?13:46
lisitsynwhy not13:46
sanujlisitsyn, i actually had a question13:46
@wikinglisitsyn, did you have an option?13:46
@wikinglisitsyn, to do it or not?13:46
lisitsyngo to army13:46
@wikinglisitsyn, i had no option....13:47
sanujlisitsyn, oh that happens in russia13:47
sanujis it like israel?13:47
Saurabh7wiking: yo13:47
sanujlike you have to serve in the army13:47
@wikingSaurabh7, so what with knn13:47
lisitsynyou have to but you can avoid that13:47
@wikingi took much longer than expected :)13:47
@wikingSaurabh7, but anyhow if this is green13:47
@wikingand i can dig through the code13:48
@wikingcan u look at the next task we discussed?13:48
sanujlisitsyn, i'm considering masters in US, but it's damn costly13:48
@wikingsanuj, i would *never* pay for education13:48
@wikingget a scholarship13:48
sanujnot good grades13:48
Saurabh7wiking: yes, I removed most external code, should be ok now13:48
sanujis there another way to get scholarship?13:48
Saurabh7i will send that on top of this13:49
sanujPhD can be funded :D13:49
lisitsynoh introducing other policy to any13:49
lisitsynbroke too much13:49
sanujlisitsyn, ohkay13:49
lisitsynI am still struggling with SG_ADD13:49
lisitsynbut its close13:50
sanujlisitsyn, we have this in swig13:50
sanuj%template(TagKernel) Tag<CKernel*>;13:50
sanujlisitsyn, i have forgotten why we decided on CKernel* :D13:50
sanuji mean why the pointer13:50
lisitsynbecause we have to stick with pointers13:50
lisitsynall the typemaps are working with pointers13:50
sanuji remember something like, i tried CKernel but it was not working13:51
sanujso moved to pointers13:51
sanujbut i don't remember exactly what13:51
lisitsynit can't be CKernel13:51
lisitsynit should be either Some<CKernel> or CKernel*13:52
lisitsynotherwise nothing is shared13:52
sanujlisitsyn, sets("gaussian", kernel) also works instead of setsKernel("gaussian", kernel)13:53
sanuji checked it on local13:53
sanujbut this would break the symmetry in sets(), gets(), has()13:53
lisitsynyou mean type is changed?13:54
sanujwiking went crazy after i told him that our api is like setsX("name", X_obj)13:54
lisitsynwell that's the only way13:55
sanujso i tried sets("name", X_obj) and it also works13:55
lisitsynI get what you mean by symmetry13:55
sanujlisitsyn, but gets("name") won't work unless you find a better implementation13:55
sanujlisitsyn, i mean, type info is saved in the Any object13:56
sanujcan't we recast it without passing the type info via getsX()?13:56
lisitsynit could be possibly done in python13:57
lisitsynmaybe wiking's favourite js13:57
lisitsynbut no java and no c++13:57
sanujlisitsyn, we will have to save the type info in the map in SGObject::Self13:57
sanujbut i guess that is not desired13:57
lisitsynit is not about saving the type13:58
lisitsynyou just can't have variable return type13:58
lisitsyn<??> gets()13:58
sanujlisitsyn, c++ limits us in this way13:59
lisitsynwell you can stay in Any type as long as you can13:59
lisitsynit can be done13:59
lisitsynspeaking of symmetry13:59
lisitsynmay be it is ok to get a little bit asymmetrical13:59
lisitsynwe set much more often than we get13:59
sanujso shall i change it?14:00
lisitsynI don't mind14:00
lisitsynprobably it is better14:00
sanujbut then usage of TagX in sets() would be totally redundant14:00
sanujobj.sets(tag_x, obj_x)14:01
sanujobj.sets("name_x", obj_x)14:01
sanujeveryone will use the latter one14:01
sanujwho will make a Tag object14:01
lisitsynno reason except performance14:02
sanujlisitsyn, is Tag more efficient?14:02
lisitsynwell you don't compute hash14:03
sanujoh yeah14:03
sanujlisitsyn, cool, i'll keep both14:03
sanujlisitsyn, shall we do the next steps in plugins after GSoC ends?14:04
lisitsynyeah maybe14:04
sanuji'll focus on readmes now14:04
sanujlisitsyn, one more thing14:15
sanujshall i move this:14:15
sanuj#ifdef HAVE_JINJA214:15
sanuj%include "shogun-base.i"14:15
sanujfrom SGBase.i to modshogun.i?14:15
lisitsynno idea14:16
sanujbecause then14:16
sanuji can replace statements like this14:16
sanuj%template(TagRealVector) Tag<SGVector<float64_t>*>;14:16
sanuj%template(TagRealVector) Tag<RealVector*>;14:16
sanujas RealVector is defined in Library.i14:17
sanujbut Library.i is not visible in SGBase.i14:17
lisitsynmove it then14:17
sanujlisitsyn, i'm moving it back to SGBase.i14:27
sanujbecause it still gives the same error14:27
-!- sanuj [~sanuj@] has quit [Ping timeout: 250 seconds]15:27
@wikinglisitsyn, around?17:43
-!- shogun-notifier- [] has joined #shogun18:37
shogun-notifier-shogun: OXPHOS :feature/cereal * 42b5a51 / / (10 files):
shogun-notifier-shogun: cereal serialization - SGVector/Any/SGObject18:37
shogun-notifier-shogun: Viktor Gal :feature/cereal * 898282e / / (10 files):
shogun-notifier-shogun: Merge pull request #3375 from OXPHOS/cereal_kmeans18:37
shogun-notifier-shogun: Cereal serialization - SGObejct18:37
-!- sanuj [~sanuj@] has joined #shogun19:03
-!- sanuj [~sanuj@] has quit [Remote host closed the connection]19:37
-!- shogun-notifier- [] has quit [Quit: transmission timeout]21:37
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/] has quit [Quit: - A hand crafted IRC client]22:22
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/] has joined #shogun22:23
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/] has quit [Quit: - A hand crafted IRC client]22:36
--- Log closed Sat Aug 06 00:00:53 2016