IRC logs of #shogun for Thursday, 2016-05-12

« Wednesday, 2016-05-11 Index Friday, 2016-05-13 »
--- Log opened Thu May 12 00:00:52 2016
-!- ptizoom [~christian@host-92-21-193-241.as13285.net] has joined #shogun00:04
-!- ptizoom [~christian@host-92-21-193-241.as13285.net] has quit [Ping timeout: 246 seconds]00:09
-!- ptizoom [~christian@host-92-21-204-114.as13285.net] has joined #shogun00:23
-!- ptizoom [~christian@host-92-21-204-114.as13285.net] has quit [Ping timeout: 240 seconds]00:30
-!- travis-ci [~travis-ci@ec2-54-147-166-192.compute-1.amazonaws.com] has joined #shogun00:43
travis-ciit's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/12952074800:43
-!- travis-ci [~travis-ci@ec2-54-147-166-192.compute-1.amazonaws.com] has left #shogun []00:43
-!- ptizoom [~christian@host-92-21-208-34.as13285.net] has joined #shogun01:18
-!- lambday [6d94a322@gateway/web/freenode/ip.109.148.163.34] has joined #shogun01:43
-!- mode/#shogun [+o lambday] by ChanServ01:43
-!- lambday [6d94a322@gateway/web/freenode/ip.109.148.163.34] has quit []01:52
Saurabh7thats a one long discussion04:30
Saurabh7Hopefuuly I have setup this irssi properly from your recommendations04:31
-!- mizari [~mizari@95-174-213-100.nts.su] has joined #shogun06:27
-!- sonne|osx [~sonne@x4e33fe28.dyn.telefonica.de] has joined #shogun07:20
-!- sonne|osx [~sonne@x4e33fe28.dyn.telefonica.de] has quit [Client Quit]07:21
@wikingarianepaola: ok so i've looked around again08:03
@wikingthis seems to be the best setup.py for me https://github.com/telamonian/ocudump/blob/master/setup.py08:04
@wikingto start with that does something with cmake08:04
@wikinglemme know if you have a better idea08:04
-!- besser82 [~besser82@fedora/besser82] has joined #shogun09:50
-!- mode/#shogun [+o besser82] by ChanServ09:50
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 276 seconds]10:13
-!- HeikoS [~heiko@host-92-0-162-192.as43234.net] has joined #shogun10:59
-!- mode/#shogun [+o HeikoS] by ChanServ10:59
CaBamorning11:03
@wikingy011:04
-!- HeikoS [~heiko@host-92-0-162-192.as43234.net] has quit [Ping timeout: 265 seconds]11:10
-!- HeikoS [~heiko@nat-232-56.internal.eduroam.ucl.ac.uk] has joined #shogun13:19
-!- mode/#shogun [+o HeikoS] by ChanServ13:19
-!- HeikoS [~heiko@nat-232-56.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 276 seconds]13:55
CaBaafter using ParameterCombination::apply_to_machine(m) - is there a way to get the parameter combination out of m again?14:30
CaBaprobably not, since apply_to_machine() distributes the parameters around where they belong...14:36
CaBa*sigh*14:40
CaBaI begin to think every class in shogun should have an obtain_from_generic() cast function :P14:40
Saurabh7CaBa: what do you mean by get out? values are already there in ParameterCombination object15:20
Saurabh7where else did you need a cast15:21
CaBaSaurabh7: in my case i didn't have the parametercombination object later on15:26
CaBaSaurabh7: oh it's just a bit annoying that in python i can't do anything like 'crossval.get_machine().get_C1()' or so, because all the getters return superclass pointers and the casting can't happen in the py-world...15:27
Saurabh7CaBa: you can get the individual paramters form the machine I think , and they should be the right type ideally15:29
CaBaSaurabh7: yes, but i have to keep a properly typed reference to the machine15:30
Saurabh7if they are not, the getters from the machine shoudl definelty return right type if you ahve an example do tell15:30
Saurabh7so it shoul dbe fixed15:30
Saurabh7CaBa:oh i see you are trying to get it from the crossval obj15:45
-!- mizari [~mizari@95-174-213-100.nts.su] has quit [Quit: Leaving]15:46
CaBaSaurabh7: yes. a lot of the getter functions that can be found in the api doc are pretty much useless for users who access them through the swig python equivalents...15:51
CaBawhat happened to CrossValidation::set_conf_int_alpha()?16:07
arianepaolawiking: thanks16:15
@wikingasdf16:23
-!- HeikoS [~heiko@nat-242-194.internal.eduroam.ucl.ac.uk] has joined #shogun16:35
-!- mode/#shogun [+o HeikoS] by ChanServ16:35
@wikingkoko HeikoS16:42
@HeikoSjo16:43
-!- HeikoS [~heiko@nat-242-194.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]17:07
-!- HeikoS [~heiko@nat-242-194.internal.eduroam.ucl.ac.uk] has joined #shogun17:08
-!- mode/#shogun [+o HeikoS] by ChanServ17:08
-!- ptizoom [~christian@host-92-21-208-34.as13285.net] has quit [Ping timeout: 260 seconds]17:22
-!- ptizoom [~christian@host-92-21-235-55.as13285.net] has joined #shogun17:35
-!- OXPHOS [8ca3fe9e@gateway/web/freenode/ip.140.163.254.158] has joined #shogun17:36
OXPHOSHey HeikoS, wiking! I have the summary/questions ready in my proposal, on page3-4.17:41
OXPHOSCan you guys have a look if you have time? Thanks!17:41
OXPHOShttps://docs.google.com/document/d/1eozTFX_mgKx3eXQfQnP07Rxi8GYKfGYdu_qvGpLRnjw/edit?usp=sharing17:41
@HeikoSOXPHOS: checking17:41
@HeikoSOXPHOS: commented a bit17:46
@HeikoSmaybe email rahul the link so he can comment as well17:46
OXPHOSHeikoS: Sure thanks. So basically the last idea is what we want to apply (If SGMatrix::onGPU() == false && linalg::have_backend(GPU) == true)17:48
OXPHOSHeikoS: No #ifdef is needed17:48
-!- travis-ci [~travis-ci@ec2-54-161-113-158.compute-1.amazonaws.com] has joined #shogun17:48
travis-ciit's lambday's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/12975833317:48
-!- travis-ci [~travis-ci@ec2-54-161-113-158.compute-1.amazonaws.com] has left #shogun []17:48
@HeikoSyes exactly17:52
@HeikoSno ifdef needed17:52
@HeikoSand the gpu_linalg falls back to the cpu one if the SGMatrix doesnt live in GPU memory17:52
@HeikoSwiking: ^17:52
OXPHOSGot it17:54
@wikingyep18:02
@wikingok i'll check the docs18:02
@wikingarianepaola: do you know packer?18:02
OXPHOSAnd thanks @ lisitsyn for showing me some.h yesterday.. Didn't realize smart_pointer is already there!18:05
lisitsynya18:05
lisitsynyes under sophisticated name18:05
lisitsynOXPHOS: there is also 'Unique'18:06
@HeikoSOXPHOS: did you check cereal already?18:08
OXPHOSlisitsyn: Haha I'm too late18:08
OXPHOSHeikoS: Yes I checked the docs and played around for a bit18:08
OXPHOSHeikoS: But haven't made any progress from yesterday yet18:09
@HeikoSkk18:09
@HeikoScurious how it will play with the smart pointers18:09
@HeikoSwiking: any news on GPL-BSD?18:09
@HeikoSI noticed a renaming18:10
@wikingHeikoS: i'll ping agian18:10
@wikingbecause i have no fucking idea18:10
@wikingwhere he disappeared18:10
@HeikoSkk18:10
@HeikoSits soon time to get that out18:10
@HeikoSshould be next step for us18:11
@HeikoSwiking: btw we still have this other option18:15
@HeikoSof releasing a bsd archive with no history18:15
@HeikoSand only bsd code18:15
@HeikoSthat we just decided to make bsd since we own the code18:15
@HeikoSand that doesnt contain any gpl code18:15
@wikingthat's not really an option18:18
@wikingbecause where do you store that code?18:18
@wikingi mean you want to generated that when there's a release?18:18
-!- besser82 [~besser82@fedora/besser82] has joined #shogun18:18
-!- mode/#shogun [+o besser82] by ChanServ18:18
@wikingchange license header etc?18:18
@HeikoSwiking: for this single release18:26
@HeikoSyes18:26
@wikingbut this release18:26
@wikingis not about bsd18:26
@wiking:D18:26
@wikingor is it?18:26
@wikingit is18:26
@wiking:)18:26
@wikingwell yeah dunno18:26
@HeikoSthis release was supposed to be the last nonc++11 one18:26
@wikingpinged my frient18:26
@HeikoSand we wanted to make it bsd as well18:26
@wiking*firend18:26
@wiking*friend18:26
@HeikoSso that we have one non c++11 AND bsd18:26
@wikingbut w8ing still18:26
@wikingyep yeo18:26
@wikingi remember now18:26
@wikingdunno what to say18:26
@wikingit's all fishy18:27
@wikingthe problem is that if we fuckup18:27
@HeikoSI can give up on that bsd thing18:27
@wikingit's all void18:27
@HeikoSyes18:27
@HeikoSbut if you want it really badly that there is a bsd shogun out there before gsoc18:27
@HeikoSwe can do that archive18:27
@HeikoSas a "special edition" ;)18:27
@wiking:>18:27
@HeikoScould do a fork and hand hack it18:27
@HeikoSbut well18:27
@HeikoSnot ideal18:27
@wikingfork ssssssssssnooogood18:28
@wikinganyhow18:28
@wikinglets wait till next monday18:28
@HeikoSok18:28
@wikingand fire up aws18:28
@wiking:D18:28
@HeikoShehe ;)18:29
@HeikoSjust worried that gsoc start soon18:29
@HeikoSand OXPHOS needs the smart pointers for cereal18:29
@wikingyeah18:29
@HeikoSoh man I am looking forward to dropping the serialization code we have ;)18:29
@HeikoSand to the plugins18:29
@wikingsmart pointer for cereal?18:29
@wikingyou mean c++11 for cerela18:29
@wiking:>18:29
@HeikoSyeah18:29
@wikingok who needs aws access18:30
@HeikoSdoesnt like raw ones18:30
@wikingraise your hands18:30
@HeikoSI dont really want, to not break stuff ;)18:30
@wikinglisitsyn: ?18:30
@wikingsonney2k_: ?18:30
@HeikoSSaurabh7: hi18:31
lisitsynwiking: yesp leas e18:31
lisitsynjust in case18:31
@wikingsent an email18:33
@wikingif you want passwd18:33
@wikinglemme know18:33
@wiking:>18:33
@wikingmmm18:34
@wikingforget the ssh key18:34
-!- travis-ci [~travis-ci@ec2-54-146-3-221.compute-1.amazonaws.com] has joined #shogun18:35
travis-ciit's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/12976480318:35
-!- travis-ci [~travis-ci@ec2-54-146-3-221.compute-1.amazonaws.com] has left #shogun []18:35
@wikingshould we actually have dc/os running over this shit? :)18:36
@wikingi mean for shogun-cloud18:36
Saurabh7HeikoS: yo18:38
Saurabh7the talk was pretty cool18:38
@wikingarianepaola: ^ what ya think?18:38
@wikingwhich talk? :)18:38
Saurabh7cache locality stuff which improveed performance18:39
@HeikoSSaurabh7: talk?18:39
@HeikoSah I see18:39
@HeikoSyeah18:39
@HeikoSthat might be good to check for the algos you touch18:39
@HeikoSand a report would be good for the project summary18:39
OXPHOSHeikoS: Hey I'm kinda confused why you said there's no GPUVector structure?18:40
@HeikoSOXPHOS: we decided to ditch it18:40
@HeikoSand do everything with SGMatrix18:40
@HeikoSand SGMatrix::isOnGPU() knows18:40
OXPHOSNo SGVector either?18:40
@wikingSaurabh7: where?18:41
Saurabh7wiking: what performance? i think they tried on mmd test, i will try on some of our algos18:42
OXPHOSHeikoS: No SGVector either? Or just transfer SGVector to GPUMatrix?18:43
@HeikoSOXPHOS: nono18:43
@HeikoSthere is not GPUMatrix18:44
@HeikoSbut you can do18:44
@HeikoScpu_to_gpu(SGMatrix)18:44
@HeikoSwhich does the transfer and changes a flag inside SGMatrix18:44
Saurabh7wiking: btw what profiler where you suggesting, perf right ?18:45
@wikingyep18:48
@wikinglisitsyn: got it?18:49
lisitsynwiking: yes18:49
OXPHOSHeikoS: aha this is so-called 'merge SGMatrix and GPUMatrix'18:55
@HeikoSOXPHOS: yes exactly18:55
@HeikoSno need for a new structure18:55
@HeikoSsince we are not doing compile time method calls anymore18:55
@HeikoSlike currently18:55
OXPHOSHeikoS: But sorry it made me even more confused...We use SGMatrix do to calculation on GPU?18:56
OXPHOSWhat kind of flag are you referring to?18:56
OXPHOSisOnGPU()18:57
OXPHOSHeikoS: Okay back to the first question: We use SGMatrix do to calculation on GPU? with ViennaCL?18:58
@HeikoSOXPHOS: yes18:59
@HeikoSwiking: ^18:59
@HeikoSright?18:59
@HeikoSit is just a wrapper for memory18:59
@HeikoSand knows where that memory lies18:59
OXPHOSWe explicitly transfer memory from CPU to GPU, and do the calculation there with ViennaCL library.19:00
@HeikoSyes19:00
@HeikoSpoint is: explicit transfer19:00
OXPHOSHeikoS: Aha. Unfortunately I know nothing about the memory transfer/GPU calculation yet..I'll do some research. Any suggestion where I can start or just google it?19:02
OXPHOSI'll know so much more by the end of GSoC..19:02
@HeikoSOXPHOS: I think you should study the existing calls that happen on the GPU19:04
@HeikoSthere is already transfer calls in Shogun19:05
@HeikoSwiking: any news on the IO bug?19:05
OXPHOSHeikoS: Got it. Thanks!19:06
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 260 seconds]19:14
@HeikoSOXPHOS: ask for help if things dont work19:15
@HeikoSOXPHOS: it is also new stuff for us19:15
OXPHOSHeikoS: Sure thanks!19:16
-!- besser82 [~besser82@fedora/besser82] has joined #shogun19:18
-!- mode/#shogun [+o besser82] by ChanServ19:18
@wikingwhat what?19:32
@wikingwas out for dinner19:32
@wikingHeikoS: sorry dcos today19:33
@wiking:)19:33
@wikinggotta setup something to show for aws ppl19:33
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 260 seconds]19:42
-!- Saurabh7 [Saurabh7@gateway/shell/panicbnc/x-kbrvwyydnyokpxuj] has quit [Ping timeout: 260 seconds]20:15
-!- Saurabh7 [Saurabh7@gateway/shell/panicbnc/x-sselotqfitbansqv] has joined #shogun20:16
-!- HeikoS [~heiko@nat-242-194.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 265 seconds]20:35
-!- sanuj [0e8bc402@gateway/web/freenode/ip.14.139.196.2] has joined #shogun20:50
sanujlua gave me a hard time today :P20:51
@wiking:<20:51
sanuji'm going home day after tomorrow :)20:51
sanujafter that i would be able to give loads of time to shogun :D20:52
-!- sanuj [0e8bc402@gateway/web/freenode/ip.14.139.196.2] has quit [Ping timeout: 250 seconds]20:56
-!- OXPHOS [8ca3fe9e@gateway/web/freenode/ip.140.163.254.158] has quit [Quit: Page closed]20:58
arianepaolawiking: I have heard of it21:23
-!- sanuj [0e8bc402@gateway/web/freenode/ip.14.139.196.2] has joined #shogun21:33
sanujanyone uses lua here? I have a lua doubt :/21:34
sanujregarding garbage collector21:34
-!- sanuj [0e8bc402@gateway/web/freenode/ip.14.139.196.2] has quit [Quit: Page closed]22:09
@wikingarianepaola: ok i'll prepare some AMIs with packer.io22:12
-!- travis-ci [~travis-ci@ec2-54-146-252-153.compute-1.amazonaws.com] has joined #shogun22:15
travis-ciit's lambday's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/12981547122:15
-!- travis-ci [~travis-ci@ec2-54-146-252-153.compute-1.amazonaws.com] has left #shogun []22:15
arianepaolawiking: ok22:16
--- Log closed Fri May 13 00:00:53 2016
« Wednesday, 2016-05-11 Index Friday, 2016-05-13 »

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