Open in new window / Try shogun cloud
--- Log opened Wed Nov 16 00:00:18 2016
-!- abhinav_ [~quassel@] has joined #shogun02:43
-!- abhinav_ [~quassel@] has quit [Ping timeout: 268 seconds]05:44
-!- abhinav_ [~quassel@] has joined #shogun05:44
-!- abhinav_ [~quassel@] has quit [Ping timeout: 256 seconds]06:12
-!- abhinav_ [~quassel@] has joined #shogun07:04
-!- sonne|work [~sonnenbu@] has joined #shogun09:01
-!- praisethemoon [~praisethe@] has joined #shogun09:04
-!- praisethemoon [~praisethe@] has quit [Changing host]09:05
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has joined #shogun09:05
-!- abhinav_ [~quassel@] has quit [Read error: Connection reset by peer]09:13
-!- abhinav_ [~quassel@] has joined #shogun09:14
lisitsynwiking: pluskid wrote a paper with hardt bengio et al09:21
lisitsynimpressive :)09:21
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has quit [Ping timeout: 268 seconds]10:15
-!- praisethemoon [~praisethe@] has joined #shogun10:17
-!- praisethemoon [~praisethe@] has quit [Ping timeout: 268 seconds]10:24
-!- praisethemoon_ [~praisethe@] has joined #shogun10:24
-!- praisethemoon_ is now known as praisethemoon10:25
-!- praisethemoon [~praisethe@] has quit [Changing host]10:25
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has joined #shogun10:25
sonne|worklisitsyn: did you expect anything less?10:47
lisitsynsonne|work: didn't like the paper tho10:47
lisitsynhe attacks holy vapnik10:47
sonne|worklisitsyn: do you have the url/pdf ?10:47
sonne|worklisitsyn: attacking a god smells like immediate punishment ;)10:48
lisitsynsonne|work: you can't just attack the dark emperor10:48
lisitsynit's careless10:48
sonne|workit is10:48
lisitsynsonne|work: tldr: neural nets overfit like crazy on random data and VC and Rademacher have no idea why it doesn't overfit when data is real10:49
lisitsynthe section 5 is really strange10:50
lisitsynthey reinvent KRR and say it works10:51
lisitsynUHMM OK :D10:51
praisethemoonHello everyone10:56
praisethemoonI have been having a look at shogun lately, and seems very interesting :)10:56
praisethemoonYou guys are doing great job.10:56
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has quit [Quit: Leaving]11:01
-!- praisethemoon [~praisethe@] has joined #shogun11:05
-!- praisethemoon [~praisethe@] has quit [Changing host]11:05
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has joined #shogun11:05
sonne|worklisitsyn: maybe because rademacher and VC only draw conclusions for all distributions not ones of natural property..11:27
sonne|workanyways even SVMs with gaussian kernels have VCdim == infty so...11:28
-!- abhinav_ [~quassel@] has quit [Ping timeout: 250 seconds]11:32
-!- praisethemoon_ [~praisethe@] has joined #shogun12:09
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has quit [Ping timeout: 265 seconds]12:13
-!- praisethemoon_ is now known as praisethemoon12:17
-!- praisethemoon [~praisethe@] has quit [Changing host]12:17
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has joined #shogun12:17
-!- abhinav_ [~quassel@] has joined #shogun13:13
-!- Hieronymous [~sudarshan@] has joined #shogun13:21
-!- HeikoS [] has joined #shogun13:32
-!- mode/#shogun [+o HeikoS] by ChanServ13:32
@HeikoSwiking: jo!13:35
-!- Hieronymous [~sudarshan@] has quit [Remote host closed the connection]13:46
-!- Hieronymous [~sudarshan@] has joined #shogun13:48
HieronymousHeikoS: jo!14:24
HieronymousI am going through the PCA notebook.14:25
HieronymousBut how to I create the cookbook ? I did read the docs, but its different for each type of learning problem.14:26
@HeikoSI dont understand your question14:29
HieronymousHeikoS: Never mind, I got it. Any C++ beginner task available ?14:32
HieronymousPerhaps something in linalg ? I read the wiki, it is a bit "nascent".14:33
@HeikoSYeah you could remove some warnings14:36
@HeikoScheck our buildbot page14:36
@HeikoSand the warnings for libshogun14:36
@HeikoSneed to be all resolved ;)14:37
@HeikoSsame for python14:37
@HeikoSbut I mean there is tons14:37
@HeikoScheck entrance tasks on our issues list14:37
@HeikoSthere is quite a few c++ ones14:37
@HeikoSlinalg needs more operations in the feature branch14:37
@HeikoScheck the commit history of the feature branch to get a feeling14:37
@HeikoSeventially we want all operations in SGVector to be in linalg14:38
HieronymousYeah, I am actually browsing through the entrance tasks. Nice I have something to research then. Great.14:38
-!- abhinav_ [~quassel@] has quit [Remote host closed the connection]14:40
@HeikoSHieronymous: yeah I mean there is so much stuff14:40
@HeikoSjust really pick anything that makes sense to you14:40
@HeikoSthe 5.1 milestone are the most pressing ones14:40
@HeikoSthere is some entrance ones in there as well14:40
HieronymousHeikoS: There are two feature branches for linalg , linalg_refactor and linalg_second_take. Which one should I be checking ?14:41
@HeikoSrefactor I think14:41
@HeikoSthe one OXPHOS has been active on14:41
@HeikoScheck the pull requests for that branch14:41
@HeikoSthey give you an idea of what is has gone through14:41
@HeikoSbut that is messy14:41
@HeikoSrather wait a bit with that :)14:41
-!- lambday [6a331336@gateway/web/freenode/ip.] has joined #shogun15:18
-!- mode/#shogun [+o lambday] by ChanServ15:18
@lambdayHeikoS: yo15:18
@HeikoSlambday: jo man!15:18
@HeikoSgood to see you here :)15:18
@lambdayHeikoS: yeah :)15:18
@lambdayHeikoS: will spend some time on bigtesting now15:19
-!- abhinav [~quassel@] has joined #shogun15:19
@lambdayHeikoS: BTW, the GPU kernels we have written may need revision15:19
@lambdaywas looking at sum max, etc..15:19
@lambdaythe max work group size is hardcoded.. it should be queried from the h/w15:20
@lambdayanyway will do that later15:20
@lambdaymaybe Pan has some ideas15:20
@lambdayfor now, making bigtesting merge ready :)15:20
@HeikoSyep absolutely15:20
@HeikoSbut have to do that *after* linalg is merged15:20
@HeikoSsame as for bigtest :)15:21
@HeikoScurrently, there is nothing15:21
@HeikoSbefore we do fancy things, we want at least something :)15:21
@HeikoSlet me know how it goes and whether you need help15:21
@lambdayHeikoS: so let me know if I miss anything in the task list (1) make supplying a custom kernel easier, get rid of the extra info message (2) pre-initialize things with DummeFeatures.. the user is not expected to supply dummy featureds when he's providing a custom kernel (3) anything else?15:21
@lambdaywill add the documentations next15:21
@HeikoSyeah thats good15:22
@HeikoSbut (1) is: merge15:22
@HeikoSbefore any of this is addressed15:22
@lambdayif some particular parameter is mandatory, it should be properly documented15:22
@HeikoSyeah true15:22
@lambdayHeikoS: merge it today you mean?15:22
@lambdayHeikoS: if you have some time to spare we can do that today15:22
@HeikoSno I mean merge before doing any of these points above15:22
@lambdayI'll rebase15:22
@HeikoSI have15:22
@lambdayyeah I got that15:22
@HeikoSthe builds needs to pass15:23
@lambdayokay let me rebase15:23
@HeikoSthe notebook needs to work15:23
@lambdayand check the build15:23
@HeikoSthe hsic problem needs to be resolved15:23
@HeikoSthen docs15:23
@HeikoSthen merge15:23
@HeikoSthen your list above :)15:23
@lambdayadded all to sticky notes15:25
@HeikoSnice one :)15:25
@HeikoSwe can tweet that ;)15:25
@lambdayHeikoS: shall I remove the HSIC files or just comment the whole thing out?15:25
@HeikoSlambday: remove them15:27
@HeikoSthey are in the history15:27
@lambdayHeikoS: cool15:27
@sukeyPull Request #3543 "Replace pointer and vlen with SGVector in QDA."  opened by sudk1896 -
@lambdayHeikoS: the entire dir "statistics" should go, right? we have things in a different directory now15:32
@lambdayHeikoS: shall I merge this feature branch to my develop and then push changes?15:32
@lambdayI mean after I check things15:33
@sukeyPull Request #3543 "Replace pointer and vlen with SGVector in QDA."  synchronized by sudk1896 -
@HeikoSleave in feature branch in upstream15:33
@HeikoSjust rebase there15:33
@HeikoSand polish there15:33
@HeikoSI will merge from there15:33
@lambdayHeikoS: yeah but HSIC and those other things are not in that branch.. those have to be removed from develop itself15:34
@HeikoSI dont understand15:34
@HeikoSyou have a branch that is rebased against develop15:34
@HeikoSwhere you removed hsic15:34
@sukeyPull Request #3543 "Replace pointer and vlen with SGVector in QDA." -
@HeikoSif I merge this, hsic wil be deleted in develop15:34
@lambdayHeikoS: yeah sorry I got confused15:35
@lambdayHeikoS: so things are already deleted in my branch15:35
@lambdayBAHSIC and all are taken care of, since it does compile15:36
@HeikoSremove all files/tests/folders/notebooks for that stuff as well15:36
@HeikoSand then all good15:37
@lambdayHeikoS: hmm.. just tried rebasing now against develop.. got merge conflicts for two sample test notebook15:39
@HeikoSyep that is since we had to clean things15:39
@HeikoSbefore release15:39
@HeikoSnotebook was broken15:39
@lambdayI can replace it with the one in upstream in feature/bigtest branch, right? that one works with the new interface (will check whether it does)15:44
@lambdaynah it's not updated15:45
@lambdaywow! github is broken15:47
-!- sonne|work [~sonnenbu@] has quit [Ping timeout: 256 seconds]15:47
-!- sonne|work [] has joined #shogun15:48
@HeikoShappens every now and then :)15:50
@HeikoSlambday: didnt get your question15:50
@lambdayHeikoS: gotta fix the notebook since I got merge conflicts.. there were so many.. was wondering whether I can replace the entire notebook from somewhere manually and then make changes according to our new API by running that thing15:53
@HeikoSI think I updated the notebook quite a bit15:55
@HeikoSin feature branch15:55
@HeikoSto solve this is easy15:55
@HeikoStake the feature branch version15:55
@HeikoSand apply the latest changes from develop on top of it, they are in the history15:56
@HeikoSI think these are only very few so should be easy15:56
@HeikoS(i.e. change order of commits, dont rebase the notebook)15:56
@HeikoSlambday: ^15:56
@lambdayHeikoS: okay let me try15:57
@lambdayI took the feature branch version itself.. let me look at the latest commits in develop for that15:57
@lambdaythere is just one that you did during the last release.. will apply that patch16:02
@HeikoSand these are very minor changes I did for the release16:05
@lambdayHeikoS: umm.. it has 1,107 additions and 1,089 deletions.. if you fixed the develop one to make it work with the older code, then it might conflict with the one we have in our feature branch since the API changed16:13
@HeikoSit is the format16:14
@HeikoSyou cannot apply the patch using git16:14
@HeikoShave to do by hand16:14
@HeikoSah thats annoying16:14
@HeikoSso I suggest16:14
@HeikoSyou check the changes I did for the release16:14
@HeikoSwhich you cannot16:14
@HeikoSsince i did not make a sep patch for the format change16:15
@HeikoSmy mistake16:15
@HeikoSso then, just discard my changes in develop16:15
@HeikoSjust make sure, that the notebook runs16:15
@HeikoS(I did only that basically)16:15
@HeikoSso just take your feature branch version, and period16:15
@lambdayHeikoS: cool.. just running the notebook16:17
@lambdayHeikoS: [-   "version": "2.7.6"] [+   "version": "2.7.10"]16:19
@lambdayis this the change I should do?16:19
@sukeyPull Request #3544 "Replace pointers with SGVector"  opened by souravsingh -
@HeikoSlambday: no change16:24
@HeikoSdiscard my stuff16:24
@HeikoSit was only to make it executable again16:24
@HeikoSjust open the notebook in your own jupyter  and run it16:24
@lambdayHeikoS: okay cool..16:24
@HeikoS(feeature branch version)16:25
@lambdaythe develop version will anyway get removed since it's a different file in a new dir it seems16:25
@lambdaywe removed it in feature branch16:25
-!- praisethemoon [~praisethe@unaffiliated/praisethemoon] has quit [Ping timeout: 268 seconds]16:28
@HeikoSlambday: what do you  mean?16:32
@HeikoSI thought you had conflicts16:32
@lambdayHeikoS: conflicts for the location where the feature branch version exists (!) - maybe because it was git mv'ed?16:33
@lambdayHeikoS: fixing this notebook now16:33
@sukeyPull Request #3543 "Replace pointer and vlen with SGVector in QDA." -
HieronymousHeikoS: Take a look at the PR - > I cleaned up the QDA. Btw, tests passed.16:35
@HeikoSHieronymous: cool thanks, will check later16:35
-!- praisethemoon [~praisethe@] has joined #shogun16:59
-!- HeikoS [] has quit [Ping timeout: 258 seconds]17:05
-!- praisethemoon [~praisethe@] has quit [Quit: Leaving]17:58
Hieronymouslamday: Hey!18:10
@lambdayHieronymous: yo18:11
HieronymousCould you tell me how I could begin contrubuting to linalg ?18:11
@lambdayoh dear18:11
HieronymousI see that it is very "nascent".18:11
@lambdayumm.. we re-did all the internals and interfaces18:11
HieronymousMight be good for a noob such as myself.18:11
@lambdayTBH I gotta have a look at the current state myself, but Pan did an awesome job porting all those existing methods into the new framework18:12
@lambdaybut still there are some left.. so that might be something you can look at18:12
@lambdayHieronymous: do you have experience in OpenCL and GPU programming?18:13
Hieronymouslamday: Nope. I was reading up a bit about Eigen though.18:13
HieronymousThat seemed "intuitive"18:13
@lambdayHieronymous: that's not a requirement, just curious cause we could have used your help :)18:13
HieronymousI am willing to learn new things. Shouldn't be a problem.18:14
@lambdayHieronymous: no worries.. so you can collaborate with Pan moving the existing methods to the new interface..18:14
HieronymousWhere should I look for these "existing methods"18:14
@lambdaybut maybe wait a bit, cause some of her patches are not yet merged.. you don't really want to work on something that somebody else did already18:15
HieronymousI did read the wiki on linalg and the reasons behind its inception.18:15
@lambdayHieronymous: those are in the develop branch18:15
@lambdaycheck inside mathematics/linalg/internals18:15
@lambdayHieronymous: Pan's changes are being merged into a feature branch linalg_refactor.. so that is another place to look at to know which methods are already there18:16
Hieronymouslambday: great, seems like a good starting point. Thanks.18:16
@lambdayHieronymous: not at all :)18:17
@sukeyNew Commit "Fixed the two-sample test notebook with the latest API" to shogun-toolbox/shogun by lambday:
Hieronymouslambday: OXPHOS is working on the cholseky decomposition.19:06
HieronymousIf I had to add something like QR or SVD, any hints on how I should proceed ?19:07
-!- abhinav [~quassel@] has quit [Ping timeout: 250 seconds]19:07
Hieronymouslambday: I actually checked OXPHOS's implementation of Cholesky. All of the operations are done using Eigen.19:07
HieronymousAny ideas on how I could begin ? Just general outline any resource or implementation I should model it on.19:08
Hieronymouslambday: Like I said, just hints/guidelines :) Thanks.19:09
@lambdayHieronymous: you can add other factorization methods :)19:09
@lambdaywait let me find you one issue on Github19:10
HieronymousYeah QR/SVD. Actually there is - 252619:10
Hieronymousthat is. I looked at those, QR or SVD hasn't been attempted by anyone. I was looking at OXPHOS's implementation19:10
Hieronymousof Cholesky in linalg.19:11
@lambdayHieronymous: yeah so you can add those two19:11
HieronymousI know all operations are done by Eigen anyway.19:11
HieronymousJust that I don't understand how I should actually begin ? I am new.19:11
@lambdayHieronymous: that's what we use for our CPU backend19:11
@lambdayHieronymous: no worries..19:11
@lambdaymaybe you can spend some time learning things about Eigen ?19:12
@lambdayhow to do these things properly19:12
HieronymousActually I did. I installed Eigen and am playing around with the QR module.19:12
HieronymousI will do that then.19:12
@lambdayand to know what each operation costs.. really important if you ask me19:12
@lambdayQR can just be another method parallel to cholesky19:13
@lambdayso you can add the new method by looking and following the same format19:13
@lambdayshould get you started smoothly :)19:13
@lambdayHieronymous: it's really important to have an idea about what exactly each operation does - how much additional memory it allocates etc etc.. whenever in doubt, write a small program like this shitty one -
Hieronymouslambday: Thanks a ton! I will fiddle around with this.19:17
@lambdayhaving an idea would help us prevent writing codes that are easy to write but shit when it comes to performance19:17
Hieronymouslambday: So all the decomposition methods go into the backend specifically - >LinalgBackendEigen.h ?19:18
@lambdayHieronymous: for now, yes19:18
@lambdaywe will eventually add the GPU support19:18
@lambdaybut one thing at a time19:18
Hieronymouslambday: I don't think I will be writing anything by hand. Eigen provides QR decomposition via a bunch of different methods.19:19
HieronymousI will have to research on them then.19:19
HieronymousI have gotten to know from the chat logs etc that you're the go to guy for linalg.19:19
HieronymousHence these many questions.19:19
@lambdayHieronymous: yeah.. the suggestion was not only about QR in particular.. but about using any library methods19:20
@lambdayHieronymous: well, I am not the guy anymore :) but I'll try to help as much as I can19:20
HieronymousI will definitely research a bit. Although I would leave it in good faith that the guys who built Eigen, did their thorough research19:20
HieronymousAnd cut out some work for noobs like me.19:20
@lambdayHieronymous: if you really want to work on linalg, you should definitely have a chat with Pan (OXPHOS)19:21
HieronymousYeah, Can't seem to catch her. I am more or less here the entire day. Except when I'm sleeping, of course.19:21
@lambdayit was her project this summer.. she knows it better than I do :)19:21
Hieronymouslambday: Thanks! Will do.19:22
@lambdayHieronymous: you can tag @OXPHOS on github if you have questions on linalg19:22
@lambdaysince GSoC is over, people (including me) are not online that often :(19:23
HieronymousI will do that on #2526 then. :)19:23
@lambdayHieronymous: cool! sounds like a plan!19:23
@wikingoh mqn19:23
HieronymousI understand work can eat a lot of time. I actually did my GSoC while I was interning at a diff place19:23
@lambdayHieronymous: and thanks for your interest on this.. linalg is an exciting endeavour19:23
@wikingwe need to fix our 3.3.0 eigen integration Eigen 3.3 released!19:23
@lambdayHieronymous: yeha I know19:24
@lambdaywiking: yo! how you doin' ? :D19:24
@lambdaywhy I am not surprised19:24
@lambdaywiking: what's the fix required for that?19:24
@lambdaylisitsyn: yo!!!19:24
@wikinglocode hax19:24
@wiking*code hacks19:25
@wikingbut yeah i'll have to see now19:25
@wikingas it has been released not long ago19:25
@lambdaywiking: let me know if I can help with that19:25
@wikingand wondering which are the fixes they managed to get in19:25
@lambdaywiking: BTW what happened with our Windows build?19:25
@wikingas there were some crazy bugs in 3.3 beta releases with iterators19:25
@wikinglambday, need to fix crossval19:25
@lambdaylisitsyn: how have you been? I've been absent for a long time19:26
@wikingthat's the major failing test that is failing19:26
lisitsynlambday: me too19:26
@lambdaywiking: that uses Eige!19:26
@wikingyeah it's not about 3.319:26
@wikingas we bundle there the library19:26
@wikingi mean eigen19:26
@lambdayso it always downloads the latest version?19:26
@lambdayand it breaks the shit out19:27
@wikingwhen we bundle19:27
@wikingit's totally fixed19:27
@wikingwhich is the version being bundled19:27
@wikingcrossval is broken19:27
@lambdaylisitsyn: :D I am back to the same fucking job :(19:27
@lambdaywiking: oh!19:27
lisitsynlambday: how come?19:28
@lambdaylisitsyn: well, for now.. deciding about applying for PhD19:28
@lambdaysince this job gives me enough personal time, so, good for now19:28
@lambdaywiking: is there any unit test that I can try to reproduce this error?19:29
@lambdayanything you noted?19:29
@wikingthe crossval unit test19:29
@wikingbut the error only comes out on windo19:29
@lambdayI'll have a look then19:29
@wikingan as well19:31
@wikingardkernel is broken on osx19:31
@wikingand you did some changes19:31
@wikingin gaussian kernels19:31
@wikingthat totally fucked up the kernel normalizers19:32
@wikingso rather fix that part19:32
@wikingi think Heiko has opened an issue for that bug19:33
@lambdaywiking: I see.. yeah I made the distance precomputable19:34
@lambdaycouldn't check on OSX19:34
@wikingyep and it does not comply with19:34
@wikingthe normallizer thing19:35
@wikingis reproducable anywhere19:35
@lambdaywiking: shall I look at our buildbots then?19:35
@wikingjust check for the issue19:35
@wikingit'll explain what's the problem19:35
@lambdaywiking: cool.. let me check19:35
@lambdaywait let me read19:36
@lambdayI am assigning it to myself19:36
@lambdaywill send a fix19:36
@sukeyIssue #3504 "fix kernel normalisers for GaussianKernel" assigned to: lambday by lambday -
@lambdaywiking: when are we planning for 5.1?19:38
@lambdayany dates?19:38
@wikingsooner the better19:38
@wikingbut i think due to some serious bugs19:38
@wikingblocking us for osx release19:38
@wikingwe are gonna have 5.0.119:38
@wikingas soon as we have those fixes19:38
@wikingif everything goes ok that could be already during this weekend19:38
@lambdaytell me about our versioning convention again.. x.0.1 is a bugfix release?19:39
@lambdayx.1 is a major release?19:39
@wikingit's not our convension19:39
@lambdayso bugfix is just a patch19:40
@wikingbut this patch19:40
@wikingthat you need to do19:40
-!- travis-ci [] has joined #shogun19:40
travis-ciit's lambday's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun:
-!- travis-ci [] has left #shogun []19:40
@wikingmight need to be under major19:40
@wikingbecause the api should change19:40
@wikingas you should remove using friend19:41
@wikingi mean of course19:41
@wikingdepends how you are planning to fix this whole story19:41
@lambdaywiking: I guess when we merge my bigtest feature branch, it should also go under a major release.. cause the API changed19:43
@lambdaybut we'll discuss that later19:43
@wikingyeah i mean around jan19:43
@wikingwe'll have 6.019:43
@wikingbut first we wanna have a bugfix release19:44
@wikinglambday, add to the milestone
@lambdaywiking: cool19:44
@wikingand here's the long list of bugs needs to be closed for 5.119:44
@lambdaycrap I really have been off for a long time :(19:45
@lambdaylooks like the issues I was talking about are already added to the milestone19:46
@lambdaywiking: did we make c++11 mandatory for 5 release already?19:47
@wikingthat's 6.019:48
@lambdayshould go on Jan release then19:48
* lambday going off20:20
-!- lambday [6a331336@gateway/web/freenode/ip.] has quit []20:20
-!- Hieronymous [~sudarshan@] has quit [Quit: Leaving]20:23
-!- shogun-buildbot [] has quit [Quit: buildmaster reconfigured: bot disconnecting]21:47
-!- shogun-buildbot [] has joined #shogun21:51
-!- HeikoS [] has joined #shogun22:53
-!- mode/#shogun [+o HeikoS] by ChanServ22:53
@wikingHeikoS, new waldeck is good23:20
@HeikoSwiking: cool will check23:20
@HeikoSwiking: hahaha :D23:30
@wikingHeikoS, here?23:48
@wikingHeikoS, so they just relesed23:48
@wikingcouple of days ago 3.3.023:48
@wikingso yeah23:49
@wikingnow fixing all that shit23:49
@HeikoSwith that thing in the changelog?23:49
@wikingbut the iterator thing23:49
@wikingis fixed23:49
@wikingthnx god23:49
@HeikoSand the operator mess?23:49
@wikingi think is well23:49
@wikingbut now i have to fix osx23:49
@wikingso i will tell you once that's done23:49
@HeikoSI want to write a bit of dev docs23:49
@wikingbecause basically i have the eigen 3.3.0 locally23:49
@HeikoSI spend too much time epxlaining to the kids23:49
@wikingso i'm doing both at the same time23:49
@wikingyeah i see that23:50
@wikinglet them write it23:50
@HeikoStakes too much time23:50
@HeikoSproof reading and feedbacking it23:50
@HeikoSI will do very minimal23:50
@HeikoSsince I agree with lisitsyn that docs are controversial as they outdate fast23:50
@HeikoSand wrong docs is even worse than no docs ;)23:50
@HeikoSbut yeah will put some time into that23:50
lisitsynTRUE THAT23:50
lisitsynwiking: you know what?23:51
lisitsynI know you're python fan23:51
lisitsynso well23:51
@wikingfuck YOU23:51
@wikingand fuck alllllllllllllll this z generation23:51
@wikingwhere this is the norm23:51
@wikinglisitsyn, ^23:52
@HeikoSwiking: btw today i played a  bit with tensorflow23:53
@wikingyeah this is like the perfect picture of the state of the art23:53
@HeikoSit is so much nicer than theano23:53
@HeikoSyet very very slow if you just write down what you think it should be23:53
@HeikoSI might start a little mini lib for phd23:53
@HeikoSand then learn more about it that way23:53
@HeikoSso we can copy andor use good bits23:54
lisitsynwiking: soo23:54
lisitsynthe thing is23:54
lisitsyna few weeks ago I wrote a mail here at my job23:54
lisitsynlike lets start writing java23:54
lisitsynit was for team of ~2023:55
lisitsynand we actually started23:55
lisitsynand now I am constantly having fun23:55
lisitsynwiking: for example a few guys were making a mock23:56
lisitsynbecause they feel anxious about their code23:56
lisitsynfailing in the night23:56
lisitsynbecause of typo23:56
lisitsynhow fucking come we get there?23:56
lisitsynis it any sane :D23:56
@wikingyeah man23:57
@wikingthat is fucking insane23:57
@wikingspelling error23:57
@wikingand maybe it's just in the catch part of the try block23:57
@wikingso it only gets there when there's the exception23:57
lisitsynexcept NameError23:57
@wikingand then suddenly like booom23:57
lisitsynok ok it wokrs23:57
@wikingno such variable23:57
@wikingbecause you actually did a spelling mistake or refactor mistake23:57
@wikingand did not change the variable part23:58
@wikingin the except part of the thing23:58
@wikingand then you are like23:58
@wikingWTF is this place?23:58
@wikingas a friend told me one day23:58
@wiking"i'm too old to learn dynamic languages"23:58
lisitsynso I am actually moving from23:59
lisitsynlike getting used to python23:59
@wikingyeh me too23:59
@wikingi loved it23:59
@wikinglike 3 years ago23:59
lisitsynto having fun of python people23:59
@wikingi was like this is great23:59
@wikingbut now i'm like23:59
@wikingare you fucking serious:)23:59
--- Log closed Thu Nov 17 00:00:01 2016