IRC logs of #shogun for Tuesday, 2017-06-20

« Monday, 2017-06-19 Index Wednesday, 2017-06-21 »
--- Log opened Tue Jun 20 00:00:21 2017
--- Log closed Tue Jun 20 02:15:50 2017
--- Log opened Tue Jun 20 02:22:02 2017
-!- shogun-toolbox [~shogun@7nn.de] has joined #shogun02:22
-!- Irssi: #shogun: Total of 21 nicks [4 ops, 0 halfops, 0 voices, 17 normal]02:22
-!- Irssi: Join to #shogun was synced in 7 secs02:22
-!- mikeling|afk [uid89706@gateway/web/irccloud.com/x-nwhagerlzhenkeyg] has joined #shogun03:02
-!- mikeling|afk is now known as mikeling03:15
shogun-buildbotbuild #282 of trusty - libshogun - viennacl is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/282  blamelist: MikeLing <sabergeass@gmail.com>, Viktor Gal <vigsterkr@gmail.com>03:15
shogun-buildbotbuild #239 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/239  blamelist: MikeLing <sabergeass@gmail.com>, Viktor Gal <vigsterkr@gmail.com>03:38
@wikingshogun-buildbot: force build --branch=develop 'deb1 - libshogun'03:41
shogun-buildbotbuild #240 forced03:41
shogun-buildbotI'll give a shout when the build finishes03:41
shogun-buildbotHey! build deb1 - libshogun #240 is complete: Failure [failed compile]03:42
shogun-buildbotBuild details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/24003:42
@wikingshogun-buildbot: force build --branch=develop 'deb1 - libshogun'04:57
shogun-buildbotbuild #241 forced04:57
shogun-buildbotI'll give a shout when the build finishes04:57
shogun-buildbotbuild #241 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/24105:01
@wikingshogun-buildbot: force build --branch=develop 'deb3 - modular_interfaces'05:04
shogun-buildbotbuild #214 forced05:04
shogun-buildbotI'll give a shout when the build finishes05:04
shogun-buildbotbuild #214 of deb3 - modular_interfaces is complete: Failure [failed configure]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/21405:05
@wikingshogun-buildbot: force build --branch=develop 'deb3 - modular_interfaces'05:11
shogun-buildbotbuild #215 forced05:11
shogun-buildbotI'll give a shout when the build finishes05:11
shogun-buildbotHey! build deb3 - modular_interfaces #215 is complete: Failure [failed configure]05:11
shogun-buildbotBuild details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/21505:11
@wikingshogun-buildbot: force build --branch=develop 'deb3 - modular_interfaces'05:12
shogun-buildbotbuild #216 forced05:12
shogun-buildbotI'll give a shout when the build finishes05:12
shogun-buildbotHey! build deb3 - modular_interfaces #216 is complete: Failure [failed install]06:04
shogun-buildbotBuild details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/21606:04
@wikingshogun-buildbot: force build --branch=develop 'deb3 - modular_interfaces'06:53
shogun-buildbotbuild #217 forced06:53
shogun-buildbotI'll give a shout when the build finishes06:53
shogun-buildbotHey! build deb3 - modular_interfaces #217 is complete: Failure [failed examples and unit tests]07:07
shogun-buildbotBuild details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/21707:07
mikelingwiking: hi, what should I do next?08:22
mikelingfor dynamicArray08:22
@wikingjust a sec08:22
mikelingok08:22
@wikingor like 2 minutes rather08:22
@wikingok08:56
@wikingi have a question08:56
@wikingmikeling,08:56
@wikingdo you know where these ctors are being called08:56
@wikinghttps://github.com/shogun-toolbox/shogun/pull/3832/files#diff-1a43b9169d617ac582b2a21332230795R10108:56
@wikinghttps://github.com/shogun-toolbox/shogun/pull/3832/files#diff-1a43b9169d617ac582b2a21332230795R12508:56
mikelingwiking: no, I don't know08:57
@wikingmmm could you create a list08:57
@wikingwhere these ctors are being used08:57
@wikingplz08:57
mikelingwiking: ok08:58
@wikingbecause acutally in case08:59
@wikingstd::vector08:59
@wikingbool p_free_array, bool p_copy_array08:59
@wikingare pretty problematic08:59
mikelingmmm, you mean we actually don't need them?09:00
mikelingfor std::vector09:00
@wikingwell you always copy the values09:01
@wikinghence you always free the copied values09:01
@wikingbut not the passed T*09:01
@wikingso yeah09:05
@wikingi was hoping to figure out where are those apis used09:05
@wikingand try to eliminate it09:05
@wikingif possible09:05
@wikingmikeling, easiest way to figure it out is to delete those ctors and try to compile :)09:08
mikelingwiking: I just add SG_PRINT in theose ctors, and run  unit test09:09
@wikingmmmm09:09
mikelinghope that could works :)09:09
@wikingbetter delete the ctors09:10
@wikingbecause it might not be09:10
@wikingpart of a unit test09:10
@wikingas not everything is tested09:10
@wikingin shogun unfortunately09:10
mikelingmmm, ok09:10
-!- geektoni [c1cdd24b@gateway/web/freenode/ip.193.205.210.75] has joined #shogun09:17
@wikingmicmn, ping09:21
micmnwiking: here09:31
@wikingmicmn, i'm wondering which pr is mergable?09:32
@wikingi know msvc is failing09:32
@wikingbut that's another story09:32
micmnlinalg ops are ok, I need to push some minor things and there are a few comments from Heiko but I don't think they apply directly to the pr09:35
micmnbut he commented here https://github.com/shogun-toolbox/shogun/pull/3842#pullrequestreview-4461805509:36
micmnfor LDA see my last comment https://github.com/shogun-toolbox/shogun/pull/3826#issuecomment-30939176009:37
micmnKernelPCA it's ok too, the only thing is the method apply_to_string_features() that I'm not sure how we should handle https://github.com/shogun-toolbox/shogun/pull/3842#pullrequestreview-4479187509:38
micmnah, Heiko created the brach feature/linalg_untemplated but I can't push to it right?09:40
micmn*branch09:40
@wikingyeah no09:40
@wikingyou can PR to that09:40
@wiking:P09:40
micmnok!09:41
mikelingwiking: here is place I found where these ctors are being called https://gist.github.com/MikeLing/1d4decc84505e0215c95796ce4bd48ed09:53
mikelingboth of bool p_free_array and bool p_copy_array been set as true09:54
@wikingok so it's always copy09:56
@wikingand always free09:56
mikelingAnd I think we could just delete these two ctors and remove p_free_array, p_copy_array09:56
mikelingyes09:56
@wikingyes09:56
@wikingthat would be the first step09:56
@wikingbut dont push09:57
mikelingok09:57
@wikingi'll give you a gist for how to solve temporarily the serialization09:57
mikelingthank you09:57
@wikingmikeling, https://gist.github.com/vigsterkr/6b557ad7f4b23280980f7507e79cdbad10:07
@wikingok10:07
@wikinglemme change10:07
@wiking2 things10:09
@wikingname of m_array_content maybe not the best10:09
@wikingyou can remove load_serializable_pre10:09
@wikingand there these changes should be marked TODO or FIXME10:09
@wikingso that shows that these are temporal workaroudns10:10
mikelingwhy we use load_serializable_post instead of load_serializable_pre?10:10
-!- TingMiao [uid229534@gateway/web/irccloud.com/x-ufgcvkdtqvyqbein] has joined #shogun10:13
-!- iglesiasg [~iglesiasg@217.119.234.214] has joined #shogun10:17
-!- mode/#shogun [+o iglesiasg] by ChanServ10:17
@wikingmikeling, because now this is a trick we do10:18
@wikingbasically say that m_array_content is something that the parameter framework can handle10:18
@wikingand seriealize10:18
@wikingso we use that as a dummy variable when serializing10:18
@wikingbut not other times10:18
@wikingso what we do is that save_serializable_pre we fill that dummy variable with the values of the std::vector10:19
@wikinglet the parameter framework serialize the values10:19
@wikingthen when load_serializable called you first make sure that everything is loaded to the class correctly10:19
@wikingthen with the post hook (load_serializable_post) you basically make sure that the values of this dummy variable is copied to the std::Vector and free that memory (as it's copied)10:20
@wikingwhen load_serializable_pre is called as the name suggest10:20
@wikingdeserialization (loading of values) is not available yet10:20
mikelingwiking: ok, I see.10:22
mikelingthank you!10:22
mikelingso the next step is to make Parameter can  deal with add(std::vector) the std::vector<T> or ?10:25
@wikingmmm10:27
@wikingwe can ignore that10:27
@wikingbecause we gonna be sooner or later10:27
@wikingthrow out the whole Paramater class10:27
@wikingso it's ok10:27
@wikingthis is just a temporal workaroudn10:27
mikelingok :)10:28
@wikingbtw10:29
@wikingwe should maybe discuss some10:30
@wikingthe story of capacity, reserve10:30
@wikingmikeling, because actually that's what DynamicArray is mimicing10:31
@wikingoriginally10:31
@wikingso for example here10:32
@wikinghttps://github.com/shogun-toolbox/shogun/pull/3832/files#diff-1a43b9169d617ac582b2a21332230795R4410:32
@wikingone should do rather10:32
@wikingm_array.reserve(resize_granularity)10:33
@wikingand then10:34
@wiking                inline int32_t get_array_size()10:34
@wikingwould actually return10:35
@wikingcapacity()10:35
@wikingm_array. capacity()10:35
@wikinginstead of m_array.size()10:35
@wikingmikeling, see what i mean?10:35
mikelingyes10:35
mikelingBTW, this free https://gist.github.com/vigsterkr/6b557ad7f4b23280980f7507e79cdbad#file-dynamicarray-h-L510:35
mikelingsomehow error out10:36
@wiking?10:36
@wikingwhere?10:36
mikelinglike https://pastebin.mozilla.org/902536810:36
@wikingwhen?10:36
@wikingoh ok lemme see10:37
@wikingbecause i only ran DynamicArray unit tests10:37
@wikingso i guess there's somewhere else10:37
mikelingmmm, but the error is for CDynamicArrayTest/0.save_serializable test10:39
mikelingI just comment the SG_FREE and build it again10:39
@wiking?10:40
@wikingthat didn't fail for me10:41
@wikingi mean w/o the sg_free it has to work10:41
@wikingi'm really wondering why would it fail with sg_free10:41
mikelingmmm, I guess it's my problem. Let me do some test10:45
mikelingI will ping you if I figure it out10:45
@wikingok but is it failing even w/o free?10:45
@wikingbtw have you removed load_serializable_pre10:45
@wiking?10:45
mikelingyes10:46
@wiking ok but is it failing even w/o free?10:46
@wiking ok but is it failing even w/o free?10:46
@wiking?10:46
-!- iglesiasg [~iglesiasg@217.119.234.214] has quit [Read error: Connection reset by peer]11:09
-!- iglesiasg [~iglesiasg@217.119.234.214] has joined #shogun11:22
-!- mode/#shogun [+o iglesiasg] by ChanServ11:22
-!- geektoni [c1cdd24b@gateway/web/freenode/ip.193.205.210.75] has quit [Quit: Page closed]11:29
-!- HeikoS [~heiko@89.105.104.229] has joined #shogun11:42
-!- mode/#shogun [+o HeikoS] by ChanServ11:42
@wikingHeikoS, i wanna throw out Parameter.cpp now11:46
@wikingthis is a fucking shit that holds us back from many imporvements11:46
@HeikoSwiking: ok11:46
@wikingand it is broken11:46
@wikingi mean our serialization11:46
@wikingis broken11:46
@HeikoSthere is two options11:46
@wikingmostly11:46
@wikingwe are now fucking aroudn here with DynamicArray11:46
@wikingpost/pre11:46
@HeikoS1.) temporarily disable clone equals and serialization11:46
@wikingbullshit11:46
@HeikoS2.) first fix tags11:46
@HeikoSwiking: I am OK with 111:47
@HeikoSjust remove all that11:47
@HeikoSso we dont have clone/equals/save for a while in develop11:47
@wikingit's so fucking tiresome to try things get work11:47
@HeikoSyes11:47
@HeikoSdo it11:47
@wikingwith this horrendous stuff11:47
@HeikoSI can help11:47
@HeikoSsome algos will break11:47
@HeikoSGP as well11:47
@HeikoS(gp modelselection(11:47
@HeikoSand model selectio11:47
@HeikoSbut yes do it11:47
@HeikoSmake an issue with the list of things to make working again11:47
@wikingHeikoS, https://gist.github.com/vigsterkr/6b557ad7f4b23280980f7507e79cdbad11:48
@HeikoSbut you are right we have to break11:48
@wikingthis is a temporal fix around11:48
@wikingthe std::vector11:48
@wikingbut still it breaks11:48
@wikingdunno why11:48
@wikingand ont the other hand11:48
@wikingsince in swig std::vector is mapped in every language11:48
@wikingwe wouldn't even need DynamicArray in the fist place11:48
@wiking*first11:48
@HeikoSyes11:48
@HeikoSwell start a branch11:48
@wikingif we would support parameter->add(std::vector)11:48
@HeikoSand try to make it build11:48
@HeikoSno11:48
@HeikoSit doesnt help11:49
@HeikoSonly equals will work11:49
@HeikoSthe rest wont11:49
@HeikoSclone/save11:49
@HeikoSI think make a branch, delete Parameter.cpp and then work the way back in11:49
@HeikoSdisabling methods (and putting those in a TODO list)11:49
@HeikoSand then work down the list, making things work with tags11:49
@wikingfuck11:49
@HeikoSnot too hard11:49
@wikingi dont even understand11:49
@wikingwhy11:49
@HeikoSequals just reads memory11:50
@HeikoSso thats ok11:50
@wikinghttps://gist.github.com/vigsterkr/6b557ad7f4b23280980f7507e79cdbad#file-dynamicarray-h-L4511:50
@wikingif i change this11:50
@wikingto11:50
@HeikoSbut anything that allocates will not work without extra modifications if you add add(std::vector)11:50
@wikingSGVector<T> m_array_content;11:50
@wikingand use11:50
@wikingSG_ADD(11:50
@wiking                            &m_array_content, "m_array_content", "whether array must be freed",11:50
@wiking                            MS_NOT_AVAILABLE);11:50
@wikingall's good11:50
@wikingif i use11:50
@wikingthat one in gist11:51
@wikingi have some random segfaults11:51
@wikingwtf?11:51
@HeikoSI think you miht have to do some static casting there11:51
@HeikoSit might call the wrong method11:51
@HeikoSpointer of pointer stuff11:51
@HeikoSbut dont know11:51
@wikinganyhow that workaroudn should work11:52
@wikingof course it has the sideeffect11:52
@wikingthat when deserializing11:52
@wikingyou have a double memory until std::vector is not set11:52
@HeikoSywp11:53
@wikingbut i mean fuck this11:53
@HeikoSthere will also be problems when scaling this up to SGObject11:53
@wikinghorrible11:53
@wiking?11:53
@wikingwhat do yo umean scaling up?11:53
@HeikoSyou dont want to add this method to every class that has a std::vector member11:53
@wikingah i wanna give up11:53
@wikingmmm11:54
@wikingwe need a wrapper11:54
@wikingin the fucking parameter11:54
@wikingas otherwise this is just bull fucking shit11:54
@wikingah i'm tired of this code11:54
@HeikoShave to switch the type in base class :D11:54
@wikingi have a feeling11:54
@HeikoSwiking: ok lets stop moaning then11:54
@HeikoSstart a branch11:54
@HeikoSdelete the class11:54
@wikingthat sometimes11:54
@wikingstarting a new code11:54
@HeikoSlets see what is left after11:54
@wikingwould be easier11:54
@wikingthan trying to stuff this somehow11:54
@wikingto make it work11:54
@HeikoSyes, I think it is ok top drop some features11:55
@HeikoSok I gotta go now11:55
@HeikoSttly11:55
@wikingHeikoS, oh yeah totally unrelevant12:59
@wikingping me when you are back12:59
-!- HeikoS [~heiko@89.105.104.229] has quit [Quit: Leaving.]14:04
-!- geektoni [~geektoni@151.62.57.216] has joined #shogun14:42
geektoniping wiking15:17
@wikingpong15:17
geektonihey, my PRs fail every time on Travis because of its time limit :/15:18
geektonilike this one https://travis-ci.org/shogun-toolbox/shogun/jobs/24468888515:18
geektonishould we add RxCpp to the Docker container?15:18
geektonibecause it may be the reason behind this problem :P15:18
@wikingwe can15:18
geektonikk, I'll do it right now ;)15:19
geektoniwiking: how is it going with AppVeyor? ;)15:20
@wikingprobably have to change parallel build to single threaded build15:21
@wikingwill take way longer :(15:21
@wikingbut should solve the problems15:21
geektoni:( what a pain!15:22
lisitsynwiking: geektoni: you guys around?15:42
geektonilisitsyn: I'm here :)15:47
lisitsynrxcpp15:55
lisitsynis it the must-have dependency now?15:55
geektoniexactly15:56
-!- iglesiasg [~iglesiasg@217.119.234.214] has quit [Quit: leaving]17:12
-!- geektoni [~geektoni@151.62.57.216] has quit [Quit: Leaving.]18:08
-!- OXPHOS_ [92bd305b@gateway/web/freenode/ip.146.189.48.91] has joined #shogun18:12
mikelingwhat's the meaning of call init_shogun() before using the library, dying.18:35
mikelinglisitsyn: ^18:35
-!- OXPHOS_ [92bd305b@gateway/web/freenode/ip.146.189.48.91] has quit [Quit: Page closed]20:30
--- Log closed Wed Jun 21 00:00:22 2017
« Monday, 2017-06-19 Index Wednesday, 2017-06-21 »

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!