Open in new window / Try shogun cloud
--- Log opened Tue Dec 15 00:00:29 2015
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun05:25
shogun-notifier-shogun: Wu Lin :develop * a2a0629 / src/shogun/ (4 files): https://github.com/shogun-toolbox/shogun/commit/a2a06292182a7543bb1872cb998636ad599643e205:25
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation05:25
shogun-notifier-shogun: Wu Lin :develop * 9795470 / src/shogun/ (4 files): https://github.com/shogun-toolbox/shogun/commit/979547067b9eedbaec04b082ef603e1e5dadb6fc05:25
shogun-notifier-shogun: Merge pull request #2937 from yorkerlin/develop05:25
shogun-notifier-shogun:05:25
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation05:25
shogun-buildbotbuild #1053 of nightly_default is complete: Failure [failed notebooks]  Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/1053  blamelist: Wu Lin <yorker.lin@gmail.com>, Bj?rn Esser <me@besser82.io>06:03
shogun-buildbotbuild #2729 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/2729  blamelist: Wu Lin <yorker.lin@gmail.com>06:16
shogun-buildbotbuild #80 of FC22 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC22%20-%20libshogun/builds/80  blamelist: Wu Lin <yorker.lin@gmail.com>06:19
shogun-buildbotbuild #1095 of FCRH - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/1095  blamelist: Wu Lin <yorker.lin@gmail.com>06:20
-!- HeikoS [~heiko@98.143.210.167] has joined #shogun06:26
-!- mode/#shogun [+o HeikoS] by ChanServ06:26
shogun-buildbotbuild #2718 of deb3 - modular_interfaces is complete: Failure [failed csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/2718  blamelist: Wu Lin <yorker.lin@gmail.com>06:31
-!- travis-ci [~travis-ci@ec2-54-147-130-225.compute-1.amazonaws.com] has joined #shogun06:37
travis-ciit's Wu Lin'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/9690249006:37
-!- travis-ci [~travis-ci@ec2-54-147-130-225.compute-1.amazonaws.com] has left #shogun []06:37
@HeikoSwiking: yoyo06:38
shogun-buildbotbuild #694 of deb4 - python3 is complete: Failure [failed test python modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb4%20-%20python3/builds/694  blamelist: Wu Lin <yorker.lin@gmail.com>06:42
shogun-notifier-shogun: Heiko Strathmann :develop * 5af68ed / / (11 files): https://github.com/shogun-toolbox/shogun/commit/5af68edff0f7ea7100c142d280307a0e7acaf35006:44
shogun-notifier-shogun: remove migration methods from all classes and blacklist depending libshogun examples06:44
shogun-notifier-shogun: Sergey Lisitsyn :develop * d33ab5a / CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/d33ab5a464df684f92c426000969aa6fabe87d2f06:44
shogun-notifier-shogun: Make integration testing an option and disable it by default06:44
shogun-notifier-shogun: Heiko Strathmann :develop * c6f9d2f / / (12 files): https://github.com/shogun-toolbox/shogun/commit/c6f9d2f2df14b0014a175212254ad293bb13fcdd06:44
shogun-notifier-shogun: Merge pull request #2938 from shogun-toolbox/feature/remove_migration06:44
shogun-notifier-shogun:06:44
shogun-notifier-shogun: Remove migration and disable integration testing06:44
wikingHeikoS: you still in canada?06:44
wikingwhat benchmarks?06:48
@HeikoSwiking: yes in canada06:58
shogun-buildbotbuild #382 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/382  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>07:00
wikingso what benchmark/07:01
shogun-buildbotbuild #383 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/383  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com>07:02
@HeikoSwiking: I just referenced you to one07:03
@HeikoSwiking: would be cool to maintain these like the unit tests07:03
@HeikoSwiking: so that we can at any point check how fast things are, etc. Helps discussions when we forgot why we decided things07:04
@HeikoSwiking: and how are things anyway?07:04
@HeikoSwiking: want to come to London in spring too?07:04
wikingmmm07:04
wikingbut wait07:04
wikingso fuck07:04
wikingwhy do we do this again07:05
wikingthat we roll our own benchmark07:05
wikinghttp://www.mlpack.org/benchmark.html07:05
wikinghttps://github.com/szilard/benchm-ml07:05
wikinglet's try to use something that then we can objectively compare07:05
wikingto other stuff out there07:05
@HeikoSwiking: sure07:06
@HeikoSwiking: for comparing high level that is better07:06
@HeikoSwiking: but we have internal design decisions07:06
@HeikoSlike how we do certain operations07:06
wikingyeah i know that07:06
@HeikoSthis is just to measure the speed of matrix products etc07:06
wikingbut i really think07:06
wikingthat having first the highline thing07:06
wikingis better07:06
@HeikoShow much to gain with GPU over CPU etc07:06
wikingthan fucking around with speed of matrix07:06
@HeikoShighline thing?07:06
@HeikoSah yeah07:06
@HeikoSagreed07:06
wikingfirst lets see why our kmeans sucks07:07
@HeikoSwiking: well lambday wrote the thing already07:07
wikinglet see why we are slow in the algo level07:07
wikingi know07:07
@HeikoSwould be stupid to loose it07:07
wikingi just dont see why is it good :D07:07
@HeikoSwiking: yeah totally right you are07:07
wikingbecause essentialy07:07
wikingyou can do this07:07
wikingwith a good profiler07:07
wikinglike perf07:07
wikingso no need for low level shit07:07
wikingthat's my 2 cents07:07
@HeikoSwiking: ok then07:07
wikingwe need high level understanding07:07
wikingfirst07:07
wikinghave it automated07:07
wikingand have it for all the methods07:08
@HeikoSwiking: how would you go ahead07:08
wikingand actually on tengible amount of data07:08
wikingwell either of those07:08
wikingcould be a good starter07:08
wikingbecause they already have integration points07:08
@HeikoSits actually on my list to do such a thing, but dont know how: check all algorithm implementations and make sure they are fast07:08
wikingwith other libs07:08
@HeikoSjust going through mlpacks benchmarks you think?07:08
wikingthat's why it's better to use something that is out there07:08
wikingyeah that could be one option07:08
wikingor the other... szilard07:08
wikingi mean that shit07:09
wikingif we cannot beat those07:09
wikinglibraries07:09
@HeikoSwiking: +107:09
wikingwe are definitely doing something very wrong07:09
wikingall of them are like java, python r based shit07:09
@HeikoSwiking: we could write a priority list of algorithms where we want to be fast07:09
wikingso you can imagine the performance07:09
wikingso if we are doing our job right07:09
@HeikoSwiking: its quite hard to beat well written python code in C to be honest and I assume thats the same for other languages07:09
wikingthen we could at least beat those algos07:09
wikingby a factor of 1007:09
@HeikoSwiking: any systematic idea?07:09
wikingdoh?07:09
@HeikoSto approach that?07:09
wikingpython07:09
wikingwhere the fucntion call07:10
@HeikoSwiking: I mean ML code07:10
wikingis a significant overhead? :D07:10
wikingcome on... python itself is a shit07:10
@HeikoSwhere there are few function calls, mostly liunalg07:10
wikingwhen it's about performance07:10
wikingyou have to use C extension in python07:10
wikingdo have performance07:10
wikingbut then the memory is always pumped up07:10
wikingbecause it's python07:10
@HeikoSagreed07:10
@HeikoSwiking: that would be a fun thing to work on together at a code sprint07:11
wikingyeah ah07:11
wikingso i wanted to reply07:11
wikingi'm gonna be in europe07:11
@HeikoSwhen?07:11
wikingbetween march and end of may07:11
wikingso any time between that time07:11
wikingwould be optimal for me07:11
wikingto join the sprint07:11
wikingi'm gonna be in wien07:11
wikingso i can jump over to london anytime07:11
@HeikoShey thats great07:12
@HeikoSso lambday is in london from 01.03 to 31.05 at least07:12
@HeikoSlisitsyn:  joins hopefully in march07:12
@HeikoSif you come, we are 407:12
wikingok just tell me the days07:13
wikingand i'm there07:13
@HeikoSI talked to s?ren who is already interested07:13
wikingbut fuck london? :)07:13
shogun-buildbotbuild #3510 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3510  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>07:13
wikingcant we find a more expensive place? :D07:13
wikingor we rent a whole flat07:13
wikingfor the group?07:13
wikinga pint of beer is soooo overpriced :)07:14
wikingand the coffee too07:14
wiking:)07:14
-!- HeikoS1 [~heiko@98.143.210.167] has joined #shogun07:14
HeikoS1ah here I am again07:14
HeikoS1wiking: last msg was fuck london07:14
wikingjust sent you the history07:15
wikingvia mail07:15
wiking:D07:15
-!- kcm [7ab1100e@gateway/web/cgi-irc/kiwiirc.com/ip.122.177.16.14] has joined #shogun07:15
-!- HeikoS [~heiko@98.143.210.167] has quit [Read error: No route to host]07:15
wikingbasically that london is $$07:15
wikingcan we get UCL to give as a hack space?07:15
shogun-buildbotbuild #3511 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3511  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com>07:16
HeikoS1wiking: I think yes07:16
HeikoS1will sort that out when I am back07:16
HeikoS1gatsby is in new building with lots of free space so might easily work07:17
HeikoS1can share a flat from airbnb07:17
wikingk07:17
wikinglet ssee07:17
HeikoS1or if it is just the two of you, you might be able to fit in my space07:17
wikingstill07:17
wikingpint and coffee is overpriced :D07:17
HeikoS1can use gatsby coffee machine07:17
HeikoS1new brewing equipment unfortunately ;)07:18
-!- kcm [7ab1100e@gateway/web/cgi-irc/kiwiirc.com/ip.122.177.16.14] has quit [Client Quit]07:18
wiking:D07:18
HeikoS1wiking: what about creating a list with algorithms that are slow in Shogun? in Wiki07:18
HeikoS1I guess SVM is fine07:18
HeikoS1mostly all the GSoC stuff will be slow07:18
HeikoS1because people just write it and stop once it works, never tune things07:19
wikingyeah07:19
wikinglet's see what these benchmarking thigns07:19
wikingare doing07:19
wikingwhat are the algos07:19
wikingand then add some of ours07:19
wikingand see wtf07:19
wikingi need to wrap up my thesis correction07:20
wikingthis month07:20
wikingso i'm a bit still busy07:20
wikingbut startin jan07:20
wikingi'll jump in as much as i can07:20
wiking(just defended the thesis yesterday night :P))07:20
wiking\o/07:20
HeikoS1really07:21
HeikoS1congrats!!07:21
HeikoS1didnt know07:21
HeikoS1yeah we can do this in spring then07:21
HeikoS1good topic for the sprint07:21
wiking:>07:21
HeikoS1as well as sergeys plugins07:21
wikingindeed07:21
HeikoS1and more code dropping07:21
HeikoS1I like it that the last patch deleted 1.5k lines from the base class!07:21
wiking:D07:21
wikingi'll look into this month07:22
wikingabout having parsers07:22
HeikoS1and we finally disabled this stupid time consiming integration tests07:22
wikingin our stuff07:22
HeikoS1need to unit test typemaps now instead though07:22
HeikoS1ah cool07:22
HeikoS1yea07:22
wikingbut i wanna supporrt07:22
HeikoS1would be good to be able to load data I guess ;)07:22
wikingbig files07:22
wiking:D07:22
wikingthat's why libarchive07:22
HeikoS1yes exactly07:22
HeikoS1I see that07:22
wikingso that you can directly read whatever.csv.gz07:22
HeikoS1just need to make sure we dont stop there but also actiually implement a csv reader07:23
HeikoS1or not implement07:23
HeikoS1integrate07:23
HeikoS1such things should not be in shogun07:23
wikingyeah07:23
wikingor at least07:23
HeikoS1as well as all the old data structures07:23
wikingmodulerized out07:23
HeikoS1should go07:23
wikingwith sergey's attempt07:23
HeikoS1to libraries that are used07:23
HeikoS1serialisation also07:23
HeikoS1the code is so bad07:23
wikingmmm serialization is still good07:23
wiking:)))07:23
wikingcome on07:23
wikingit's not that bad07:23
wikingit's abstracted nicely07:23
HeikoS1I think it is07:23
wikingthe shit i see nowadays07:24
HeikoS1haha07:24
wikingthis code in shogun07:24
HeikoS1well I would prefer a nice linb07:24
wikingis amazing07:24
wikinglinb?07:24
HeikoS1lib07:24
HeikoS1library07:24
wikingah yeah07:24
wikingwell07:24
HeikoS1lisitsyn wants to just have a map of parameters07:24
HeikoS1that is easy to store07:24
wikingsome things we cnanot afford07:24
HeikoS1and then you have serialisation07:24
wikingyeah07:24
wikingmap07:24
wiking:)))07:24
wikingdont get me started on that :D07:24
HeikoS1haha07:24
wikingbut yeah at least hashmap is not that expensive in c++07:24
wikinglike in jvm07:24
wiking:>07:24
HeikoS1hehe07:25
wikingbut o07:25
wikingok07:25
wikingi looked into a nice serialization stuff07:25
wikingbut i truely believe07:25
HeikoS1which one?07:25
wikingthat once we have the modularized system07:25
wikingthen we wouldnt care so much07:25
wikinghaving our own serialization07:25
wikingcerela07:25
wikingcereal07:25
wikinghttp://uscilab.github.io/cereal/07:25
HeikoS1wiking: I partly agree07:26
HeikoS1sure then such codes dont hurt us so much07:26
HeikoS1but the point is that we are a ML library07:26
wikingyeah sure i understand07:26
HeikoS1and we should keep our code focussed on that07:26
wikingbut as said07:26
wikingif it's just a module07:26
wikingwe can leave it there haning07:27
HeikoS1cerial sounds good!07:27
wikinghanging07:27
wikingand you can right away switch to another module/lib07:27
wikingif you want07:27
HeikoS1serialisation is needed though07:27
HeikoS1ah yes07:27
HeikoS1replace07:27
HeikoS1and then just get rid of old07:27
HeikoS1I agree07:27
HeikoS1good stuff07:27
HeikoS1hey how do you want to fix that csharp build?07:28
wikingmmm07:28
wikingi was already digging in that07:28
HeikoS1one way is to remove the two methods with the same signature07:28
wikingyeah no07:28
wiking:)07:28
wikingi'll try the better way07:28
wikingthe thing is that this all comes from the typemap07:28
wikingdefinitions07:28
wikingand that's why there's suddenlty 2 params with the same name07:28
wikingi just wanna change that07:28
HeikoS1wiking: as rename the parameters?07:29
wikingnever really managed to have the full 2 days07:29
wikingto fix it07:29
wikingyeah07:29
HeikoS1I want to release this week if possible07:29
HeikoS1got some momentum at the moment, and a bit of time07:29
HeikoS1sergey also helps07:29
wikingtyis weeek?!1?!07:29
wikingare you crazy? :D)07:29
wikingi thought we are doing release in jan07:29
wikingthat's what lisitsyn told me the last time07:29
HeikoS1when we met here, we kind of agreed to do now07:30
HeikoS1and then another one07:30
HeikoS1more releases needed07:30
HeikoS1minor releases07:30
HeikoS1ppl think the project is dead07:30
wikingyeah i get that07:30
wikingok then do now a release without a fix for c#07:30
wikingi mean if you use old mono07:30
wikingit'll work07:30
wiking:D07:30
HeikoS1ok07:30
HeikoS1but its that patch that broke it07:30
HeikoS1was working before07:30
HeikoS1I just dont get what exactly breaks it, everything is hidden from swig07:31
wikingnono07:31
wikingc# hasn't worked07:31
wikingwith mono 3.x07:31
wikingsince ever07:31
wiking:D07:31
wikingmaybe on travis it was working07:31
wikingsince it's 12.04 ubuntu07:31
HeikoS1yes07:31
HeikoS1check before and after07:32
HeikoS1the PR is sent07:32
shogun-notifier-shogun: Heiko Strathmann :develop * ece1370 / NEWS: https://github.com/shogun-toolbox/shogun/commit/ece13700f63f28c5e67660253813935ab5dd20e507:36
shogun-notifier-shogun: Add latest PR merge07:36
-!- travis-ci [~travis-ci@ec2-54-81-227-202.compute-1.amazonaws.com] has joined #shogun07:38
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/9691147207:38
-!- travis-ci [~travis-ci@ec2-54-81-227-202.compute-1.amazonaws.com] has left #shogun []07:38
HeikoS1wiking: ok I am going to bed07:39
HeikoS12am here07:39
HeikoS1n807:39
shogun-buildbotbuild #384 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/384  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>07:39
shogun-buildbotbuild #3512 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3512  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>07:40
-!- HeikoS1 [~heiko@98.143.210.167] has quit [Quit: Leaving.]07:47
-!- travis-ci [~travis-ci@ec2-23-22-250-109.compute-1.amazonaws.com] has joined #shogun08:50
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/9691613208:50
-!- travis-ci [~travis-ci@ec2-23-22-250-109.compute-1.amazonaws.com] has left #shogun []08:50
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]10:36
-!- sonne|work [~sonnenbu@ip5b415e18.dynamic.kabel-deutschland.de] has joined #shogun11:29
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has joined #shogun14:45
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has quit [Ping timeout: 250 seconds]14:50
-!- sanuj [~sanuj@117.204.244.140] has joined #shogun16:25
sanujanyone there?16:27
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has joined #shogun16:47
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has quit [Ping timeout: 260 seconds]16:52
sonne|worksanuj: yes :)17:09
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun17:12
shogun-notifier-shogun: yorkerlin :develop * 274a975 / src/shogun/lib/StringMap.h,src/shogun/optimization/lbfgs/LBFGSMinimizer.h: https://github.com/shogun-toolbox/shogun/commit/274a975c7b0ae879c16783027f54a398cf73319017:12
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation17:12
shogun-notifier-shogun: Wu Lin :develop * 3fce935 / src/shogun/lib/StringMap.h,src/shogun/optimization/lbfgs/LBFGSMinimizer.h: https://github.com/shogun-toolbox/shogun/commit/3fce9353e9564305cfbd05ca1921d39c935ea96317:12
shogun-notifier-shogun: Merge pull request #2941 from yorkerlin/develop17:12
shogun-notifier-shogun:17:12
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation17:12
shogun-buildbotbuild #385 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/385  blamelist: yorkerlin <yorker.lin@gmail.com>, Wu Lin <yorker.lin@gmail.com>17:14
shogun-buildbotbuild #3513 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3513  blamelist: yorkerlin <yorker.lin@gmail.com>, Wu Lin <yorker.lin@gmail.com>17:16
sanujsonne|work: hello17:17
sanuji'm installing shogun17:17
sonne|worksanuj: ok!17:17
-!- HeikoS [~heiko@98.143.210.167] has joined #shogun17:17
-!- mode/#shogun [+o HeikoS] by ChanServ17:17
sanujshall i install it using the latest release or from the develop branch?17:17
sanuji need the python bindings as well17:18
sonne|worksanuj: depends on what you need17:19
sonne|workdevelop branch should work fine as well since we will be doing a release soon17:19
sanuji plan to write a ipython notebook so i need the installation for that17:19
sonne|worksanuj: well just install it with the python bindings then...17:21
sanujsonne|work: cool, thanks :)17:21
-!- HeikoS [~heiko@98.143.210.167] has quit [Ping timeout: 260 seconds]17:22
-!- HeikoS [~heiko@modemcable027.154-203-24.mc.videotron.ca] has joined #shogun18:33
-!- mode/#shogun [+o HeikoS] by ChanServ18:33
@besser82sanuj, hey which os are you on?18:34
sanujbesser82: linux mint18:35
@HeikoSubuntu18:35
@besser82sanuj, I see...  possibly the ubuntu ppa will work on that one too18:35
sanujHeikoS: yeah18:35
@HeikoSsanuj: compile problems?18:36
@besser82sanuj, was asking, because there are distros with redy2go packages of shogun18:36
sanujHeikoS: i haven't compiled till now18:37
@besser82like I'm maintainig the Fedora rpms18:37
@HeikoSsanuj: yes, and?18:37
sanujbesser82: yeah, i know18:37
-!- travis-ci [~travis-ci@ec2-54-198-199-141.compute-1.amazonaws.com] has joined #shogun18:38
travis-ciit's Wu Lin'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/9701177118:38
-!- travis-ci [~travis-ci@ec2-54-198-199-141.compute-1.amazonaws.com] has left #shogun []18:38
sanujHeikoS: i was just asking that, should i compile from the release version or develop branch18:38
@besser82ahh, okie, you already got information18:38
@HeikoSsanuj, it depends on what you want to do18:38
@HeikoSif you want to develop, compile yourself18:38
@HeikoSotherwise, check whether things you need are in last release18:39
sanujHeikoS: write ipython notebook for now18:39
@HeikoSrelease is fine18:39
@besser82sanuj, then the ppa will be the better choice18:39
sanujHeikoS: https://github.com/shogun-toolbox/shogun/issues/270118:39
sanujbesser82: cool18:40
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has joined #shogun18:48
@HeikoSsanuj: I see18:49
@HeikoScool that you work in this18:49
sanujHeikoS: my work was mainly before last year's gsoc18:50
sanujcontinuing back now18:51
@HeikoSgreat18:53
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has quit [Ping timeout: 256 seconds]18:53
@HeikoSsanuj: you can also look at using the new Shogun linalg interface to do NNs18:53
@HeikoSin particular for convolution operations18:53
@HeikoSsince these things might soon run on GPU18:53
sanujHeikoS: i was about to ask18:54
sanujdo you guys plan to have GPU support18:54
@HeikoSwe do18:54
@HeikoScheck Convolve.h18:54
sanujcuda cudnn or something like that18:54
sanujokay18:54
@HeikoSand Convolve_unittest.cc18:56
@HeikoShow to use them18:56
@HeikoSthese things already run on GPU18:56
@HeikoSsanuj: but if you can come up with a list of operations that are needed for the NNs let us know18:57
@HeikoSAFAIK, the NN code is also single-core18:57
@HeikoSmight be good to think about multicore18:57
sanujHeikoS: okay18:57
@HeikoSsanuj: but convolution is already useful I uess ;)18:57
sanujHeikoS: indeed18:58
@HeikoSsanuj: did you ever edit some of the NN code in Shogun?18:58
@HeikoSSorry I have  a bad memory18:58
sanuji solved some todo tasks that sergey listed in the deep learning project18:59
sanujand before that18:59
sanujdid some clean up and linalg stuff18:59
sanujdocs in the beginning to get started with the contributing process18:59
sanujit was fun ;)19:00
@HeikoSsanuj: sounds good19:00
@HeikoSif you already touched linalg19:00
@HeikoSAnother important thing we need currently: linear solves in lingalg19:00
@HeikoSbut if you use the convolution interface in the exisitng NN code, that would be super helpful already19:01
sanujHeikoS: i can look into it19:02
sanujwinter breaks here19:03
@HeikoSsanuj: ha!19:03
@HeikoSyeah19:03
@HeikoSsanuj: let me know if you need help with anything19:03
@HeikoSsanuj: we really appreciate people doing things for Shogun19:04
@HeikoSsanuj: so let's make sure you dont get stuck19:04
sanujHeikoS: yes19:04
sanujshall i start by the convolution part19:04
sanuji'll have to see the code first19:05
@HeikoSsanuj: yes, have a look at that19:09
@HeikoSsanuj: if you can systematically replace all the convolution code in the NNs by the linalg interface calls, that is a huge step forward19:09
sanujHeikoS: cool19:10
sanujthere are 2 backend for conv19:11
sanujeigen3 and viennacl19:11
sanujHeikoS: I need to go now19:16
sanujwill look into it tomorrow and get back to you19:16
@HeikoSsanuj: backend can be chosen at compile time19:16
@HeikoSsanuj: that is the idea19:16
@HeikoSthat algorithm is *independent* of backend19:17
@HeikoSso one can choose19:17
@HeikoSwithout touching19:17
@HeikoSsee you later19:17
sanujHeikoS: yeah i remember19:17
sanujbye19:17
-!- sanuj [~sanuj@117.204.244.140] has quit [Quit: Leaving]19:17
wikinggood morning canada19:19
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout]20:12
@HeikoSwiking: jojo20:23
@HeikoSwiking: lets change license20:23
@HeikoSjust do it20:23
@HeikoSI think the list is small enough20:23
-!- HeikoS [~heiko@modemcable027.154-203-24.mc.videotron.ca] has quit [Quit: Leaving.]20:23
-!- HeikoS [~heiko@modemcable027.154-203-24.mc.videotron.ca] has joined #shogun20:25
-!- mode/#shogun [+o HeikoS] by ChanServ20:25
@HeikoSwiking: think thats ok?20:26
-!- yorkerlin [42c72722@gateway/web/freenode/ip.66.199.39.34] has joined #shogun20:27
@HeikoSyorkerlin: jojo20:30
yorkerlinhihi20:30
@HeikoSyorkerlin: how are the warnings going?20:30
@HeikoSwe sill have loads of them20:31
yorkerlinhalf of them are caused by the coma20:31
yorkerlinhopefully, I can fix most of them tonight20:32
@HeikoSyorkerlin: make sure to fix them using regular expressions20:34
@HeikoSnot by hand20:34
@HeikoStakes too long20:34
yorkerlinok20:35
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun20:35
shogun-notifier-shogun: Wu Lin :develop * 83bc825 / src/shogun/optimization/ (22 files): https://github.com/shogun-toolbox/shogun/commit/83bc8257ae9c4d71652de5d36156378eefadbd0720:35
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation20:35
shogun-notifier-shogun: Wu Lin :develop * 4633c5f / src/shogun/optimization/ (22 files): https://github.com/shogun-toolbox/shogun/commit/4633c5fa16e5dfc7567b434adc644206f7f23da720:35
shogun-notifier-shogun: Merge pull request #2942 from yorkerlin/develop20:35
shogun-notifier-shogun:20:35
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation20:35
-!- HeikoS [~heiko@modemcable027.154-203-24.mc.videotron.ca] has quit [Quit: Leaving.]20:37
shogun-buildbotbuild #386 of trusty - libshogun - viennacl is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/38620:39
shogun-buildbotbuild #387 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/387  blamelist: Wu Lin <yorker.lin@gmail.com>20:41
shogun-buildbotbuild #3514 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/351420:41
shogun-buildbotbuild #3515 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3515  blamelist: Wu Lin <yorker.lin@gmail.com>20:43
shogun-buildbotbuild #2730 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/2730  blamelist: Wu Lin <yorker.lin@gmail.com>20:44
-!- HeikoS [~heiko@modemcable027.154-203-24.mc.videotron.ca] has joined #shogun20:45
-!- mode/#shogun [+o HeikoS] by ChanServ20:45
shogun-buildbotbuild #1096 of FCRH - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/1096  blamelist: Wu Lin <yorker.lin@gmail.com>20:46
shogun-buildbotbuild #81 of FC22 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC22%20-%20libshogun/builds/81  blamelist: Wu Lin <yorker.lin@gmail.com>20:47
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has joined #shogun20:50
-!- HeikoS [~heiko@modemcable027.154-203-24.mc.videotron.ca] has quit [Ping timeout: 240 seconds]20:51
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has quit [Ping timeout: 272 seconds]20:55
shogun-buildbotbuild #2719 of deb3 - modular_interfaces is complete: Failure [failed csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/2719  blamelist: Wu Lin <yorker.lin@gmail.com>20:56
-!- yorkerlin [42c72722@gateway/web/freenode/ip.66.199.39.34] has quit [Quit: Page closed]21:01
shogun-buildbotbuild #695 of deb4 - python3 is complete: Failure [failed test python modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb4%20-%20python3/builds/695  blamelist: Wu Lin <yorker.lin@gmail.com>21:05
-!- yorkerlin [42c72722@gateway/web/freenode/ip.66.199.39.34] has joined #shogun21:07
yorkerlinany idea about automatic kernel learning? eg, The prox method and sprox method should work well for any kernel function, any mean function, and any likelihood.21:12
-!- yorkerlin [42c72722@gateway/web/freenode/ip.66.199.39.34] has quit [Quit: Page closed]21:15
-!- travis-ci [~travis-ci@ec2-23-22-250-109.compute-1.amazonaws.com] has joined #shogun21:17
travis-ciit's Wu Lin'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/9705917621:17
-!- travis-ci [~travis-ci@ec2-23-22-250-109.compute-1.amazonaws.com] has left #shogun []21:17
shogun-notifier-shogun: Wu Lin :develop * 3bdefa4 / src/shogun/ (7 files): https://github.com/shogun-toolbox/shogun/commit/3bdefa45c7761005c4cd4872a4ef71e83f0a562d21:30
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation21:30
shogun-notifier-shogun: Wu Lin :develop * 1e03142 / src/shogun/ (7 files): https://github.com/shogun-toolbox/shogun/commit/1e031427510057b1aaee6a27a00102b7e78feeb021:30
shogun-notifier-shogun: Merge pull request #2943 from yorkerlin/develop21:30
shogun-notifier-shogun:21:30
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation21:30
-!- HeikoS [~heiko@98.143.210.167] has joined #shogun21:31
-!- mode/#shogun [+o HeikoS] by ChanServ21:31
shogun-buildbotbuild #388 of trusty - libshogun - viennacl is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/38821:34
shogun-buildbotbuild #3516 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/351621:35
shogun-buildbotbuild #389 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/389  blamelist: Wu Lin <yorker.lin@gmail.com>21:35
shogun-buildbotbuild #3517 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3517  blamelist: Wu Lin <yorker.lin@gmail.com>21:38
shogun-buildbotbuild #2731 of bsd1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/2731  blamelist: Wu Lin <yorker.lin@gmail.com>21:38
shogun-buildbotbuild #82 of FC22 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC22%20-%20libshogun/builds/82  blamelist: Wu Lin <yorker.lin@gmail.com>21:41
shogun-buildbotbuild #1097 of FCRH - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/1097  blamelist: Wu Lin <yorker.lin@gmail.com>21:42
shogun-buildbotbuild #2720 of deb3 - modular_interfaces is complete: Failure [failed csharp modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/2720  blamelist: Wu Lin <yorker.lin@gmail.com>21:52
lisitsynHeikoS: hey21:54
@HeikoSlisitsyn:  jo21:54
lisitsynjust got home21:54
lisitsynHeikoS: I can share .vimrc with you so you can check21:54
@HeikoSlisitsyn: cool!21:54
lisitsynhttps://gist.github.com/anonymous/2fe55e42a2b60fffd96621:55
@HeikoSlisitsyn:  any idea how to replace license consistently in all files?21:55
lisitsynHeikoS: ^21:55
@HeikoSautomatically21:55
lisitsynuhmm21:55
lisitsynsed..21:55
@HeikoSlisitsyn: do you know how to do that?21:56
@HeikoSI mean remive all GPL21:56
@HeikoSbut extract names21:56
@HeikoSand replace by BSD with those names21:56
@HeikoSOR21:56
@HeikoSwe change even more21:56
@HeikoSsaying that we have Authors file21:56
@HeikoSand dont distringuish individual files anymore21:56
@HeikoSrather use git history21:56
@HeikoSgit blame21:56
@HeikoSthoughts?21:57
@HeikoSwould be easier21:57
@HeikoSif all the headers would just be BSD license21:57
lisitsynHeikoS: I actually think we don't need full license21:57
lisitsynin the file21:57
@HeikoSor even remove license21:57
@HeikoSand have only one file21:57
@HeikoSand it is "unless otherwise stated"21:57
lisitsynHeikoS: we can just replace21:58
lisitsynargh irc crap21:58
@HeikoSlisitsyn: shall we do that before the release?21:58
lisitsyn     /** ... */21:58
@HeikoS?21:58
@HeikoSah yes21:58
lisitsynwith /** This file is distributed by BSD 2-clause license. See LICENSE for more details */21:58
@HeikoSor next release better21:58
@HeikoSgood21:59
lisitsynyes21:59
lisitsynnext release21:59
@HeikoSand no more authors in the files21:59
lisitsynwell it makes sense to have some21:59
lisitsynactually people are interested in seeing their names21:59
lisitsynso I'd actually do it like21:59
@HeikoSok21:59
lisitsyn    /** This file is distributed by BSD 2-clause license. See LICENSE for more details */21:59
lisitsynoops21:59
@HeikoShow to keep that?21:59
lisitsyn  /** This file is distributed by BSD 2-clause license. See LICENSE for more details21:59
@HeikoSI mean how to extract names21:59
lisitsyn@authors: [git blame thing]22:00
lisitsyn */22:00
lisitsynjust anybody touched the file22:00
lisitsynordered by number of lines changed22:00
lisitsyn:)22:00
lisitsyneasy to do22:00
lisitsynjust assign it to me22:00
@HeikoSlisitsyn: can you add and assign yourself?22:01
lisitsynyes22:01
@HeikoSnumber of lines touced?22:01
@HeikoSno good idea22:01
@HeikoSsince it gets outdated22:01
@HeikoSjust if ever touched better?22:01
lisitsynno I mean the order22:02
shogun-buildbotbuild #696 of deb4 - python3 is complete: Failure [failed test python modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb4%20-%20python3/builds/696  blamelist: Wu Lin <yorker.lin@gmail.com>22:02
@HeikoSlisitsyn: I see22:02
@HeikoSyeah go for that!22:02
@HeikoSgood point22:02
@HeikoSWe need to be a bit careful about some files22:03
@HeikoSbut will do that when its time22:03
@HeikoSafter release22:03
lisitsynHeikoS: we have to check the full diff afterwise though22:05
@HeikoSlisitsyn: yes22:06
-!- travis-ci [~travis-ci@ec2-54-198-199-141.compute-1.amazonaws.com] has joined #shogun22:14
travis-ciit's Wu Lin'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/9707186622:14
-!- travis-ci [~travis-ci@ec2-54-198-199-141.compute-1.amazonaws.com] has left #shogun []22:14
shogun-notifier-shogun: Wu Lin :develop * 4c4d6cc / src/shogun/optimization/ (3 files): https://github.com/shogun-toolbox/shogun/commit/4c4d6ccddd41b597c25a7c8f84cbeb51d0ab802022:26
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation22:26
shogun-notifier-shogun: Wu Lin :develop * ad768ab / src/shogun/optimization/ (3 files): https://github.com/shogun-toolbox/shogun/commit/ad768abffa709abe99bc7347988d1d33826c90c522:26
shogun-notifier-shogun: Merge pull request #2946 from yorkerlin/develop22:26
shogun-notifier-shogun:22:26
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation22:26
shogun-buildbotbuild #390 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/390  blamelist: Wu Lin <yorker.lin@gmail.com>22:29
shogun-buildbotbuild #3518 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/351822:30
shogun-buildbotbuild #3519 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3519  blamelist: Wu Lin <yorker.lin@gmail.com>22:33
shogun-notifier-shogun: Wu Lin :develop * 2c135e0 / src/shogun/optimization/ (10 files): https://github.com/shogun-toolbox/shogun/commit/2c135e0e2dd7ba3b93b8ed618a0b2806d62f305322:36
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation22:36
shogun-notifier-shogun: Wu Lin :develop * 6973b1b / src/shogun/optimization/ (10 files): https://github.com/shogun-toolbox/shogun/commit/6973b1b2d63c71c60ff25105ee10378d3bb43eec22:36
shogun-notifier-shogun: Merge pull request #2947 from yorkerlin/develop22:36
shogun-notifier-shogun:22:36
shogun-notifier-shogun: Fix python_modular warnings by adding API documentation22:36
shogun-buildbotbuild #391 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/391  blamelist: Wu Lin <yorker.lin@gmail.com>22:39
shogun-buildbotbuild #3520 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/3520  blamelist: Wu Lin <yorker.lin@gmail.com>22:40
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has joined #shogun22:52
lisitsynHeikoS: ok I see wu is making moves22:52
@HeikoSlisitsyn: yes22:52
@HeikoSpushed him22:52
lisitsyncool22:52
lisitsynI think I've got some time22:52
lisitsynok22:52
lisitsynso cookboo22:52
lisitsynk22:52
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has quit [Ping timeout: 256 seconds]22:56
@HeikoSgreat22:57
@HeikoSI can help/join22:57
@HeikoSlet me know22:57
lisitsynHeikoS: ok localized the problem22:59
-!- travis-ci [~travis-ci@ec2-54-198-199-141.compute-1.amazonaws.com] has joined #shogun23:09
travis-ciit's Wu Lin'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/9708313423:09
-!- travis-ci [~travis-ci@ec2-54-198-199-141.compute-1.amazonaws.com] has left #shogun []23:09
lisitsynHeikoS: I am worried whether it would be cross-platform enough23:13
@HeikoSlisitsyn: so?23:29
@HeikoSlisitsyn: details?23:29
lisitsynHeikoS: not clear yet23:32
lisitsynHeikoS: oh there is one more issue23:38
lisitsynwe actually did it a bit wrong23:38
lisitsynit should generate examples right into the build directory23:38
@HeikoSyou have to explain me more, I did not get what you meant previously23:38
@HeikoSlisitsyn: yes that is true23:38
lisitsynHeikoS: previous issue is that we only test on mac/linux23:38
@HeikoSthough make install should also make them available23:39
lisitsynbut should work on windows too23:39
@HeikoSlisitsyn: shogun and windows?23:39
lisitsynhaha23:39
@HeikoShaha23:39
lisitsynyes23:39
lisitsynok23:39
@HeikoSwhat is not portable23:39
@HeikoSyou were using os.sep etc23:39
@HeikoSthat should be ok23:39
lisitsynyes probably23:39
lisitsynok23:39
lisitsynHeikoS: I will continue moving stuff to build tomorrow23:40
lisitsynrelative paths we had are ugly23:40
lisitsynshould be parametrized23:40
@HeikoSlisitsyn: yeah23:42
@HeikoSlisitsyn: agreed23:42
-!- travis-ci [~travis-ci@ec2-54-147-130-225.compute-1.amazonaws.com] has joined #shogun23:46
travis-ciit's Wu Lin'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/9708515123:46
-!- travis-ci [~travis-ci@ec2-54-147-130-225.compute-1.amazonaws.com] has left #shogun []23:46
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has joined #shogun23:53
-!- shaochuan [~shaochuan@c-73-158-54-228.hsd1.ca.comcast.net] has quit [Ping timeout: 240 seconds]23:57
-!- sonney2k_ is now known as sonney2k23:58
-!- mode/#shogun [+o sonney2k] by ChanServ23:58
--- Log closed Wed Dec 16 00:00:31 2015