Open in new window / Try shogun cloud
--- Log opened Thu Jul 20 00:00:03 2017
-shogun-buildbot:#shogun- Build nightly_all #19 is complete: Failure [test (failure)] -
-!- sheogorath27 [sheogorath@gateway/shell/firrre/x-ddnnnuvzbaqvgfuj] has joined #shogun04:16
@sukey[] Pull Request synchronized by MikeLing04:21
@sukey[] Pull Request synchronized by MikeLing05:12
@sukey[] Pull Request synchronized by MikeLing06:13
-!- olinguyen [18303760@gateway/web/freenode/ip.] has joined #shogun06:42
@sukey[] Pull Request merged by vigsterkr07:22
@sukey[] vigsterkr pushed 2 commits:07:22
-shogun-buildbot:#shogun- Build osx1 - libshogun #95 is complete: Failure [failed compile (failure)] -
@sukey[] Issue opened by vigsterkr07:38
-!- geektoni [] has joined #shogun09:04
-!- iglesiasg [~iglesiasg@] has joined #shogun09:15
-!- mode/#shogun [+o iglesiasg] by ChanServ09:15
micmnwiking: osx buildbot doesn't have hdf5, so I'll revert serialization to ascii and lower the equals check accuracy09:39
@wikingmicmn, or just simply do a check for hdf509:40
@wikingand when it's false09:40
@wikinglet the that test be a NOP09:40
micmnok I thought we wanted those tests in buildbot but if that's not the case I can check for hdf5 in the makefile09:41
@wikingwelll look09:42
@wikingmost of the places it tests it09:42
@wikingosx will be fixed eventually :)09:42
@wiking(to have hdf509:42
@wikingi mean the bot09:42
@sukey[] Pull Request opened by geektoni09:49
@sukey[] Pull Request opened by geektoni09:54
@sukey[] Pull Request synchronized by micmn10:15
-!- HeikoS [] has joined #shogun10:23
-!- mode/#shogun [+o HeikoS] by ChanServ10:23
-!- mikeling [uid89706@gateway/web/] has joined #shogun11:31
@sukey[] Pull Request synchronized by MikeLing11:32
@sukey[] Pull Request synchronized by geektoni11:41
micmnwiking, HeikoS: QDA serialization is failing since an SGNDArray var is not serialized, it isn't right? there is some workaround?11:49
micmn(I mean: it isn't serializable)11:50
@HeikoSmicmn: yes it is not atm11:54
@HeikoSmicmn: can we replace it with a SGMatrix?11:54
micmnok I'll give it a try11:56
-!- HeikoS [] has quit [Quit: Leaving.]12:01
-!- HeikoS [~heiko@] has joined #shogun12:59
-!- mode/#shogun [+o HeikoS] by ChanServ12:59
@HeikoSmicmn: is using SGMatrix possible?12:59
micmnmmm it's a vector of matrices so I'd say no without ugly hacks12:59
@HeikoSare they all of the same length?13:00
@HeikoSI think then you can just stack them horizonally13:00
@HeikoSi.e. columns 1-n matrix 1, columns n-2*n matrix 2 etc13:00
@HeikoShave a big matrix that has all of those13:01
@HeikoSand then you can put a method to SGMatrix that gives you a view on a submatrix (non-owning)13:01
micmnthat's what I meant with "ugly hacks" :P13:01
@HeikoSSGMatrix(big_mat.get_column_vector(idx), size, false)13:01
@HeikoSif you hide it behind API I think it is ok13:01
@HeikoSwe will soon allow for std::vector to be a member that is serializable13:02
@HeikoSthen you can just put the SGMatrix in there13:02
@HeikoSbut for now, just write a convenience method that extract the view13:02
@HeikoSI think that is OK, as long as it is modular, i.e. the algorithm code doesnt change once we make the big SGMatrix a std::vector<SGMatrix>13:03
micmnyeah with that method it would be cleaner13:03
micmnjust a small change to the algorithm13:03
@HeikoSmethod call stays the same13:12
@HeikoSget_submatrix(i) or whatever13:12
@HeikoSand then the internals of that thing change13:13
@HeikoSactually, I think it might be worth putting that in SGMatrix13:13
@HeikoSand make clear that it is non owning13:13
-!- HeikoS [~heiko@] has quit [Quit: Leaving.]13:43
@iglesiasgit's been long time, but maybe SGMatrixList could help for this13:57
micmniglesiasg: but it isn't serializable as well, right?14:09
@iglesiasgmmm I am not sure about that (so I guess no :D)14:17
@sukey[] Pull Request synchronized by HeyItsDavid15:35
Trixislets see if i fucked up the ident again (for some reason im having trouble setting my editor up to comply with the indenting in shogun)15:39
-!- olinguyen [18303760@gateway/web/freenode/ip.] has quit [Ping timeout: 260 seconds]15:46
-!- iglesiasg [~iglesiasg@] has quit [Quit: leaving]16:00
@wikingTrixis, :)16:06
@wikingTrixis, brew install clang-format :)16:06
Trixiswiking: haha fair16:11
-!- HeikoS [] has joined #shogun16:12
-!- mode/#shogun [+o HeikoS] by ChanServ16:12
Trixiswiking: also i think i figured out why i couldnt get shogun running on cluster -- hardware issues16:12
@wikinglol whaaaaaaaaat?16:13
Trixisye the login node of a year old and fairly powerful GPU cluster, which includes homes went down, hence certain sectors were not readable/writeable. #thanksdell16:15
Trixisseems to be the reason why it was throwing weird errors about libraries not being found despite beiong correctly linked/ in path16:15
-!- HeikoS [] has quit [Ping timeout: 240 seconds]16:33
-!- HeikoS [] has joined #shogun16:49
-!- mode/#shogun [+o HeikoS] by ChanServ16:49
-!- HeikoS [] has quit [Quit: Leaving.]17:03
@sukey[] Pull Request opened by micmn17:19
@sukey[] Pull Request opened by micmn17:41
-!- HeikoS [] has joined #shogun17:43
-!- mode/#shogun [+o HeikoS] by ChanServ17:44
TrixisHeikoS: right, fixed the error + the indent, style checked passed, however the build ran quite slow and one thing on travis failed due to timeout, any idea if thats an issue on my side?18:14
@HeikoSno fails to timeout is not your fault18:15
@HeikoSill restart, you ahve a link?18:15
@HeikoSthx I restarted it18:16
@sukey[] Pull Request merged by karlnapf19:11
@sukey[] New commit by karlnapf19:11
-!- geektoni [] has quit [Ping timeout: 240 seconds]19:34
-!- HeikoS [] has quit [Ping timeout: 268 seconds]20:16
@sukey[] Issue closed by vigsterkr23:24
@sukey[] Pull Request merged by vigsterkr23:24
@sukey[] vigsterkr pushed 2 commits:23:24
@sukey[] Pull Request merged by vigsterkr23:25
@sukey[] vigsterkr pushed 2 commits:23:25
-shogun-buildbot:#shogun- Build osx1 - libshogun #98 is complete: Success [build successful] -
--- Log closed Fri Jul 21 00:00:04 2017