Open in new window / Try shogun cloud
--- Log opened Thu Jul 28 00:00:41 2016
-!- shogun-notifier- [] has joined #shogun03:23
shogun-notifier-shogun: Viktor Gal :feature/javascript * f34f2ac / / (8 files):
shogun-notifier-shogun: Add Javascript modular interface build framework03:23
@wikingbuuyaaa :)03:23
shogun-buildbotbuild #1180 of nightly_default is complete: Failure [failed test notebooks]  Build details are at  blamelist: Heiko Strathmann <>, OXPHOS <>, Viktor Gal <>03:30
-!- travis-ci [] has joined #shogun03:31
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 []03:31
-!- OXPHOS [c0a906af@gateway/web/freenode/ip.] has joined #shogun04:24
@wikingOXPHOS: hey05:35
OXPHOSwiking: hey05:36
@wikingany blockers?05:48
shogun-buildbotbuild #52 of memleak - valgrind is complete: Failure [failed memory check]  Build details are at  blamelist: Heiko Strathmann <>, OXPHOS <>, Viktor Gal <>06:07
shogun-buildbotbuild #1051 of nightly_none is complete: Failure [failed compile]  Build details are at  blamelist: Heiko Strathmann <>, OXPHOS <>, Viktor Gal <>06:08
OXPHOSwiking: hopefully no..compiling06:08
-!- shogun-notifier- [] has quit [Quit: transmission timeout]06:23
OXPHOSwiking: so06:38
OXPHOSwiking: I did in init.h06:38
OXPHOSwiking: theoretically all macros are redefined right?06:39
OXPHOSwiking: I mean I can just use cereal_save instead of save across shogun right?06:39
@wiking#ifdef CEREAL_SAVE_FUNCTION_NAME06:44
@wiking#undef CEREAL_SAVE_FUNCTION_NAME06:44
@wiking#define CEREAL_SAVE_FUNCTION_NAME cereal_save06:44
@wikingi would do06:44
@wiking#ifdef CEREAL_SAVE_FUNCTION_NAME06:44
@wiking#undef CEREAL_SAVE_FUNCTION_NAME06:44
@wiking#define CEREAL_SAVE_FUNCTION_NAME cereal_save06:44
@wikingbecause you always want to define that macro06:44
OXPHOSyou're right06:45
OXPHOSI just found out I need to include the file in unittest to enable the macro06:46
OXPHOSor actually macros.hpp is undefined in unittest06:47
-!- sanuj [~sanuj@] has joined #shogun08:05
OXPHOSwiking: hey I'm still having trouble including the header file. Can you have a look at it? Thanks!09:26
OXPHOSwiking: error:
sanujlisitsyn, yo09:26
OXPHOSwiking: PR:
-!- sanuj [~sanuj@] has quit [Ping timeout: 265 seconds]10:05
@wikingOXPHOS: i would just pull the docker image10:47
@wikingand debug it in that10:47
@wikingdocker pull shogun/shogun-dev10:48
-!- wiking [~wiking@huwico/staff/wiking] has quit [Quit: leaving]12:10
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun12:11
-!- mode/#shogun [+o wiking] by ChanServ12:11
-!- wiking [~wiking@huwico/staff/wiking] has quit [Client Quit]12:11
-!- sanuj [~sanuj@] has joined #shogun12:46
-!- HeikoS [] has joined #shogun12:46
-!- mode/#shogun [+o HeikoS] by ChanServ12:46
-!- HeikoS [] has quit [Ping timeout: 250 seconds]12:51
sanujlisitsyn, yolo12:53
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun12:58
-!- mode/#shogun [+o wiking] by ChanServ12:58
-!- wiking [~wiking@huwico/staff/wiking] has quit [Read error: Connection reset by peer]13:05
-!- wiking [] has joined #shogun13:07
-!- wiking [] has quit [Quit: leaving]13:35
-!- HeikoS [] has joined #shogun13:38
-!- mode/#shogun [+o HeikoS] by ChanServ13:39
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun13:43
-!- mode/#shogun [+o wiking] by ChanServ13:43
@wikingHeikoS, yo13:44
@HeikoSwiking: jo13:45
@wikingHeikoS, have you seen the new feature branch13:45
lisitsynsanuj: yolo13:49
sanujlisitsyn, got time ??13:49
lisitsynHeikoS: yolo13:49
lisitsyna bit13:49
sanujdid you see the PR?13:49
lisitsynnoah yet13:49
sanujlisitsyn, i need to talk about what to do next?13:49
@wikinglisitsyn, the yolo is this13:50
@wikingsanuj, dont you wanna spend some time with clang-format ? :)13:51
@wikingsince the indentation of you code is usually messed up... that way you would get the autoformatting13:51
@wikinglisitsyn, javascript modular interface13:51
lisitsynok why not13:51
lisitsynwiking: nodjes?13:51
@wikingalready have a shogun-ml npm package13:51
lisitsynoh why fucking not13:51
@wikinglisitsyn, because think about it13:51
@wikinglisitsyn, runtime speed of python vs v813:52
@wikingwhat's your guess?13:52
lisitsynv8 is like13:52
@wikingso dont you think that actually it would make much more sense13:52
@wikingto do data science in JS (or ecma6) instead of python?13:52
@wikingi mean it's a scripting language13:52
@wikingeasy to prototype etc13:53
@wikingbut still the runtime is much better...13:53
@wikingand with ecmascript6 you have a lot of nice abstractions13:53
@wikingor with typescript you have even types :)13:53
@wikingHeikoS, ^13:53
@wikingi'm trying to reach out to #scijs people13:53
@wikingto have the array/features mapping defined nicely with ndarray that they are maitaining13:54
lisitsynsanuj: do tests work now?13:55
sanujlisitsyn, i have commented in the PR13:55
sanujwiking, okay13:56
@wikingsanuj, i can give you something as a starter13:56
lisitsynsanuj: I don't really see how can it compare them13:59
lisitsynthey're slightly different in implementations13:59
sanujokay, i'll remove this test13:59
sanujlisitsyn, i can take care of the rest of the tests then14:01
sanujlisitsyn,  I think we should establish a way to add plugins in shogun and think of how to start migrating.14:02
sanujlisitsyn, so what do you think?14:02
sanujlisitsyn, wiking suggested that a shogun plugin can be an entirely new repository also14:03
sanujlisitsyn, i had mailed you about this a few days before14:04
lisitsynsanuj: I don't think we need to split repositories now14:05
sanujlisitsyn, so what should be the next step here for the plugins PR?14:06
sanujdo we need anything else before we merge this?14:06
lisitsynsanuj: make tests work and fix indentation ;)14:09
@HeikoSwiking: checking14:10
@HeikoS+1 fore js14:11
@HeikoSIs that a lot of work?14:11
@wikingmmm slowly steadily14:11
@wikingnow it at least compiles14:11
@HeikoSjust wondering14:11
@wikingneeds some tests etc14:11
@wikingfor optimal version of course it needs some work14:11
@HeikoStypemaps yet missing14:11
@wiking(having the proper memory wrappers, so not to copy mem etc)14:11
@wikingwanna talk first with the #scijs guys14:12
@wikingbefore jumping on that14:12
@wikinganyhow wondering how to extend the metastuff :)14:12
@HeikoSmeta examples?14:13
@HeikoSthats easy14:13
@HeikoSjust need a new target.json14:13
@HeikoSthat is the dictionary that maps a syntax tree to the target language14:13
@HeikoSsimple text replacement14:13
@wikingok i'll try to look into it14:15
sanuji'm getting github unicorn14:15
@wikingsanuj, check status.github.com14:15
@wikingmaybe it has a hiccup now14:15
@wikingseems like14:15
@HeikoSwiking: I can help with that,14:15
@wikingHeikoS, would be cool14:15
@HeikoSbut we need typemaps first otherwise it will just fail14:15
@wikingHeikoS, i've put together an npm package even14:15
@HeikoSnice one14:16
@wikingnono it's ok14:16
@wikingi mean the typemap i'm working on14:16
@wikingbut dunno when i can finish14:16
@wikingbut imagine on the end14:16
@wikinghaving nodejs running this14:16
@wikingwith the d3js frontend14:16
@HeikoSIll do the meta ones then14:16
@wikingas that way i can even test stuff14:16
@HeikoSyeah exactly14:17
@HeikoSwe currently only do integration testing from cpp meta examples14:18
@HeikoSwould be cool to make that work for all of them14:18
@HeikoSsince then we implicitly test typemaps14:18
@HeikoSwiking: gs14:18
@HeikoSlisitsyn: jo!14:24
@wikingHeikoS, i'm there14:31
lisitsynHeikoS: jolo14:35
@HeikoSlisitsyn: jo14:35
@HeikoSso what about meta integration testing across all modular targets?14:35
@HeikoSnot just cpp14:35
@HeikoSremember the issue?14:35
-!- sanuj [~sanuj@] has quit [Ping timeout: 244 seconds]15:24
-!- sanuj [~sanuj@] has joined #shogun15:59
-!- leagoetz [] has joined #shogun16:01
OXPHOSleagoetz: hey lea! sorry I cannot reply PM somehow16:06
leagoetzoh hey16:13
leagoetzjust wanted to remind you that you need to wrap up this week, as from next week the focus should be on polishing/merging PRs16:14
leagoetzalso ... have you updated your timeline?16:14
sanujlink to the meeting:
leagoetzyes, yes sanuj, it's all there and the time to do it has come :)16:16
sanujleagoetz, sure sure16:17
OXPHOSleagoetz: sure thanks! yes I did around midterm16:18
leagoetzand it's staying like this?16:19
leagoetzthat's fine then - make sure you can achieve everything that's on there.16:19
sanujleagoetz, you work on neurons stuff? :D16:20
leagoetzyes, neurons stuff16:20
sanujleagoetz, so what exactly?16:20
sanuji like to chat with phds about their research16:21
leagoetzI try to understand what computation individual neurons are performing16:21
sanujleagoetz, how do you do it?16:21
sanujwith fMRI?16:22
lisitsynSTACK MORE LAYERS %))))16:22
sanujsome other imaging technique16:22
leagoetzno, fMRI does not have high enough resolution16:22
sanujlisitsyn, that's for deep learning16:22
lisitsynyes just stack more layers!16:22
sanujlisitsyn, haha16:22
leagoetzI do a combination of experiments and biophysical modelling16:22
sanujlisitsyn, residual networks16:22
lisitsynmore layers16:22
leagoetzwell, in some sense individual neurons also stack more layers16:22
lisitsynmore more layers16:23
sanujlisitsyn, need some time to discuss about my timeline16:23
lisitsynrelu relu layer norm16:23
* lisitsyn away16:23
lisitsynwhen does gsoc end?16:23
leagoetzinternal deadline 16 aug16:23
lisitsynsanuj: I think you have to stay a few more years to complete plugins16:23
leagoetzexternal 2316:23
lisitsynsorry about that! :D16:23
sanuji won't be able to do gsoc16:23
sanujas i won't be a student16:24
sanujlisitsyn, but i'll stay ;)16:24
leagoetzyes, nice! :)16:24
lisitsynyeah you have to16:24
sanujleagoetz, do you have a blog or something for your research?16:24
leagoetznope, too lazy16:24
leagoetzyou can check out our lab website16:24
lisitsyntrain more linear models on hadoop16:25
sanujlisitsyn, i'm telling you, deep learning will eat everything16:25
lisitsynno chance16:26
lisitsynwe gonna drop that sheatz at some point :P16:26
sanujonly the umbrella term "deep learning" will have more techniques under it16:26
lisitsynthe winter is coming16:26
lisitsynit is I feel the breeze16:27
leagoetzoh dear, it's finally summer here and you talk about winter coming16:27
sanujleagoetz, thanks for the website link16:27
lisitsynsummer, is it like +15C there in uk?16:27
lisitsynand 90% cloudy?16:28
sanujit was 40C here a few days ago16:28
sanujnot it's 28C due to monsoon16:28
leagoetzyes, exact description ;D16:28
leagoetzsummer means, I don't wear my winter coat for clycling to uni16:28
lisitsynI wouldn't be able to move if it is 40C outside16:29
leagoetzthere is only about 2 months of the year where this is possible16:29
leagoetzJuly is one of them16:29
sanujlisitsyn, haha, it's not good16:29
leagoetz28 due to monsoon is still a bit crazy16:29
sanujleagoetz, what is summer temp in london?16:30
leagoetzanything over 15 deg celsius16:30
lisitsynI've seen some british people in spain16:30
sanuji see16:30
lisitsynthey get really crazy16:30
lisitsynI guess that's how they feel about > +15C16:31
leagoetzBritish people abroad ... the only thing worse than British people in Britain ;)16:31
lisitsynI had so much fun they're watching football all the time :D16:32
sanujlisitsyn, you like football?16:32
lisitsynwell it is fun to play but not really fun to watch16:33
sanuji used to watch16:33
sanujnow stopped16:33
sanujlisitsyn, so here is the deal16:33
lisitsynyeah at some age you stop that16:33
sanuji need to open all new PRs this week16:33
lisitsynI guess because you realize you can't spend your life on such bs16:34
sanujfrom next week i can only improve them and get them merge16:34
lisitsynsanuj: uhm ok16:34
lisitsynlets merge16:34
lisitsynthis thing16:34
lisitsynmake tests work and indents!16:34
sanujyeah, i'm adding more tests16:34
sanujwill be done before tonight16:34
sanujlisitsyn, but what else do you think i can work on16:34
sanujlike open new PRs this week16:34
lisitsynsanuj: we also need to merge Jinja16:35
sanujyeah but get() bug is messy16:35
lisitsynwith swig16:35
sanujif the function has get(), i think it errors16:35
lisitsynsanuj: ok lets think of other name maybe16:36
sanujfetch is not good i guess :P16:36
sanujleagoetz, one of the brain related blogs i read sometimes,
sanujthis guy is also at UCL i guess16:37
lisitsynsanuj: set is also bad16:37
lisitsynuhmm we need better pair16:38
sanujlisitsyn, set is too common16:38
leagoetzlooks like it16:38
lisitsynleagoetz: help us16:38
sanujlisitsyn, and HeikoS suggested to hack SG_ADD16:39
lisitsynwe need alternative to set/get16:39
sanujthese are function names ^16:39
lisitsynnot good for api16:42
lisitsynI'd get mad16:42
leagoetzbut I guess make is not the best either16:43
lisitsynok we can use some suffix/prefix for set/get16:43
lisitsynaset aget16:43
lisitsynsets gets16:43
lisitsynset_ get_16:43
leagoetzdoset doget16:44
sanujlisitsyn, i think get* will error with SWIG16:44
sanujlisitsyn, will have to try16:44
lisitsynaset/aget is actually used16:45
lisitsynin clojurescript16:45
sanujlisitsyn, i will try set/get suffix and prefix names16:46
lisitsynI like setr getr most at the moment16:46
sanujlisitsyn, what about hacking SG_ADD or something else16:46
lisitsynwe need to do that16:46
lisitsynmay be I'll try to get some time on weekend16:47
sanuji think we should have an example for how to convert old shogun class to support tags-parameters16:47
lisitsynsanuj: if we have SG_ADD all of them are already converted16:47
lisitsynsanuj: sets gets is better!16:47
lisitsynlets go with sets/gets16:47
lisitsynwiking: HeikoS: ho16:48
lisitsynwe have tough decision there!16:48
lisitsynsanuj: a few reasons16:51
lisitsyn1) it would work with set<Tab>16:51
lisitsyn2) it looks like a bit of proper english16:51
sanujlisitsyn, set<Tab>16:53
sanujyes, it's proper english16:53
sanujlisitsyn, if we hack SG_ADD and merge it, will we be able to delete Parameter.h etc16:56
lisitsynnot sure16:56
-!- leagoetz [] has quit [Remote host closed the connection]16:57
sanujlisitsyn, so shall i try to change a shogun class to use tags-parameter with old parameter framework16:58
sanujand then move on to hacking sg_add16:58
lisitsynwhich shogun class?16:58
sanujany class16:58
sanujjust to see if it works (it should work)16:59
lisitsynyeah it makes sense to try16:59
-!- leagoetz [] has joined #shogun16:59
lisitsynnot sure we should commit that tho16:59
sanujyeah, won't commit16:59
sanujlisitsyn, so i'll start these two PRs tomorrow16:59
sanujdo we need to start any new PR apart from these two for tags?17:00
lisitsynprobably no17:00
sanujlisitsyn, and for plugins?17:00
sanujnew PR after the current one?17:01
lisitsynthe deadline is quite soon so lets just polish these17:01
sanujlisitsyn, okay17:01
sanujlisitsyn, and how do we add classes in the jinja template17:02
sanujfor base-shogun17:02
lisitsyndon't get it17:03
-!- leagoetz [] has quit [Ping timeout: 250 seconds]17:03
sanujlisitsyn, we need to list the base-shogun in jinja template17:04
sanujlisitsyn, i mean, we need to list the classes for base-shogun in jinja template17:05
sanujlisitsyn, so we need to decide what classes to list in that17:06
lisitsynoh ok17:06
sanujor will we merge without the full list?17:06
lisitsynyeah merge w/o full list17:06
sanujlisitsyn, thanks!17:07
sanujleaving for dinner now17:07
-!- sanuj [~sanuj@] has quit [Ping timeout: 276 seconds]17:13
-!- leagoetz [] has joined #shogun17:21
lisitsynwiking: gets/sets17:21
lisitsynhow do you like it17:21
lisitsyninstead of get/set17:21
-!- leagoetz [] has quit [Client Quit]17:22
@wikinglisitsyn, why the diff?17:25
@wikingi mean i dont really think there's an awful big diff17:25
@wikingso if it's easier to have gets/sets instead of get/set17:25
@wikingthen lets' have it like that17:25
@wikingbecause actually semantically speaking (+english grammar) dictates gets/sets :D17:26
lisitsynwiking: no actually it is get set17:28
lisitsynlike wiking set x to 317:28
lisitsynbut yeah17:28
lisitsynif it solves the problem17:28
@wikingwiking sets x to 317:28
@wikingin english :)17:28
@wikingbut if set is easier17:29
@wikingthen let's go with that17:29
@wikingeither way17:29
@wikingi dont see it as a big problem17:29
@wikingboth 'makes' sense17:29
@wikingor you can sell the idea :D17:29
-!- sanuj [~sanuj@] has joined #shogun18:14
-!- OXPHOS [c0a906af@gateway/web/freenode/ip.] has quit [Quit: Page closed]18:36
-!- shogun-notifier- [] has joined #shogun19:51
shogun-notifier-shogun: OXPHOS :feature/linalg_refactor * 56bae12 / / (6 files):
shogun-notifier-shogun: linalg-sgvector&sgmatrix-mean19:51
shogun-notifier-shogun: Heiko Strathmann :feature/linalg_refactor * 93b1cca / / (6 files):
shogun-notifier-shogun: Merge pull request #3334 from OXPHOS/linalg_mean19:51
shogun-notifier-shogun: LinalgRefactor-mean19:51
-!- HeikoS_mobile [] has joined #shogun20:07
-!- sanuj [~sanuj@] has quit [Remote host closed the connection]20:09
-!- HeikoS_mobile [] has quit [Remote host closed the connection]20:12
-!- shogun-notifier- [] has quit [Quit: transmission timeout]22:51
--- Log closed Fri Jul 29 00:00:42 2016