IRC logs of #shogun for Tuesday, 2016-03-01

« Monday, 2016-02-29 Index Wednesday, 2016-03-02 »
--- Log opened Tue Mar 01 00:00:18 2016
-!- psitae [a5e6e0df@gateway/web/freenode/ip.165.230.224.223] has quit [Ping timeout: 252 seconds]00:00
-!- HeikoS [~heiko@host-92-0-162-192.as43234.net] has joined #shogun00:38
-!- mode/#shogun [+o HeikoS] by ChanServ00:38
-!- ebadawy [~ebadawy@197.163.32.61] has quit [Ping timeout: 260 seconds]00:42
lisitsynHeikoS: done00:47
@HeikoSlisitsyn: pushed?00:47
lisitsynyes00:47
lisitsynHeikoS: didn't test modular though00:47
@HeikoSchecking00:48
@HeikoSshould work modular00:49
@HeikoSbut travis will tell :)00:49
@HeikoSif it ever start building it00:49
lisitsynI don't see irc notifies anymore00:49
lisitsyn:(00:49
lisitsynHeikoS: ok sleep time00:49
@HeikoSgn800:49
lisitsyn2:48 is a good timing00:49
lisitsynlike in these good old days00:50
@HeikoShaha00:51
@HeikoSlisitsyn: indeed00:51
@HeikoSlisitsyn: you broke the tests for python generator00:51
@HeikoSbut otherwise brilliant00:51
@HeikoSworks00:51
@HeikoSgn800:51
-!- HeikoS [~heiko@host-92-0-162-192.as43234.net] has quit [Ping timeout: 246 seconds]00:55
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 268 seconds]01:43
shogun-buildbotbuild #587 of debian wheezy - memcheck is complete: Failure [failed memory check]  Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/587  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>01:50
-!- nelson-liu [~nelsonliu@c-73-53-48-140.hsd1.wa.comcast.net] has joined #shogun02:25
-!- nelson-liu [~nelsonliu@c-73-53-48-140.hsd1.wa.comcast.net] has quit [Remote host closed the connection]02:31
-!- Rajat [73f87233@gateway/web/freenode/ip.115.248.114.51] has joined #shogun02:32
-!- Rajat [73f87233@gateway/web/freenode/ip.115.248.114.51] has quit [Client Quit]02:33
-!- nelson-liu [~nelsonliu@nelsons-macbook-pro.dyn.cs.washington.edu] has joined #shogun03:03
-!- travis-ci [~travis-ci@ec2-50-17-47-241.compute-1.amazonaws.com] has joined #shogun03:39
travis-ciit's Sergey Lisitsyn'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/11273254303:39
-!- travis-ci [~travis-ci@ec2-50-17-47-241.compute-1.amazonaws.com] has left #shogun []03:39
-!- ebadawy [~ebadawy@197.163.32.61] has joined #shogun04:52
-!- mlenthu [0e8b5206@gateway/web/freenode/ip.14.139.82.6] has quit [Ping timeout: 252 seconds]04:59
shogun-buildbotbuild #588 of debian wheezy - memcheck is complete: Failure [failed memory check]  Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/588  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>05:17
-!- ghoshbinayak [~AndChat29@117.226.184.199] has joined #shogun05:30
-!- nelson-liu [~nelsonliu@nelsons-macbook-pro.dyn.cs.washington.edu] has quit [Remote host closed the connection]05:35
-!- nelson-liu [~nelsonliu@205.175.98.95] has joined #shogun06:01
shogun-buildbotbuild #958 of nightly_all is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_all/builds/95806:41
-!- ebadawy [~ebadawy@197.163.32.61] has quit [Ping timeout: 246 seconds]06:42
-!- nelson-liu [~nelsonliu@205.175.98.95] has quit [Remote host closed the connection]06:58
-!- sanuj [0e8bc402@gateway/web/freenode/ip.14.139.196.2] has joined #shogun07:46
-!- nelson-liu [~nelsonliu@D-173-250-194-161.dhcp4.washington.edu] has joined #shogun07:56
-!- Izaya [~Eejya@103.225.100.51] has joined #shogun08:01
-!- ghoshbinayak [~AndChat29@117.226.184.199] has quit [Ping timeout: 246 seconds]08:14
sonne|work1lisitsyn: will you serve ice cold vodka for free the whole day?08:29
lisitsynsonne|work1: warm voddka!08:30
lisitsynice cold is for canadians08:30
sonne|work1hot vodka is what you in Siberia are used to... I always forget that :P08:32
-!- sanuj [0e8bc402@gateway/web/freenode/ip.14.139.196.2] has quit [Ping timeout: 252 seconds]08:39
-!- slardar [~slardar@192.241.200.146] has joined #shogun08:56
-!- besser82 [~besser82@fedora/besser82] has joined #shogun09:09
-!- mode/#shogun [+o besser82] by ChanServ09:09
-!- ghoshbinayak [~AndChat29@117.251.192.46] has joined #shogun09:51
-!- HeikoS [~heiko@host-92-0-162-192.as43234.net] has joined #shogun10:15
-!- mode/#shogun [+o HeikoS] by ChanServ10:15
sonne|work1besser82: no I won't contact mentors - HeikoS is our leader now :)10:29
@besser82sonne|work1, Okie...  =)10:30
@besser82HeikoS, when are you planning to send out the mentoring invites for GSoC?  :P10:30
@HeikoSbesser82: sonne|work1 will do soon10:31
@HeikoSbesser82: you can co mentor the plugin project?10:31
@besser82HeikoS, Yes, sir! ^^10:31
-!- HeikoS [~heiko@host-92-0-162-192.as43234.net] has quit [Quit: Leaving.]10:41
-!- nelson-liu [~nelsonliu@D-173-250-194-161.dhcp4.washington.edu] has quit [Remote host closed the connection]10:44
-!- HeikoS [~heiko@host-92-0-162-192.as43234.net] has joined #shogun10:53
-!- mode/#shogun [+o HeikoS] by ChanServ10:53
-!- ghoshbinayak [~AndChat29@117.251.192.46] has quit [Ping timeout: 260 seconds]11:09
-!- Izaya [~Eejya@103.225.100.51] has quit [Quit: Leaving]11:10
-!- Guest92392 [ibm___@nat/iiit/x-ezdaseupxeljwuxn] has joined #shogun12:11
-!- ebadawy [~ebadawy@197.163.32.61] has joined #shogun12:13
-!- anishshah [~anishshah@27.5.145.62] has joined #shogun12:30
anishshahHi guys, I'm trying to install shogun from source. When I do "./configure", I'm getting "no such file or directory". I had previously install using package manager.12:31
anishshahany help?12:31
@besser82anishshah, we're using CMake since a quite while...  to configure you'd use: `mkdir build && cd build && cmake ..`12:32
anishshahok12:33
anishshahThanks. I will try12:33
anishshah-- Could NOT find CCache (missing:  CCACHE)12:37
anishshah-- Using system's malloc12:37
anishshah-- Could NOT find MOSEK (missing:  MOSEK_DIR MOSEK_INCLUDE_DIR MOSEK_LIBRARY)12:37
anishshah-- Looking for dgemm_12:37
anishshah-- Looking for dgemm_ - found12:37
anishshah-- A library with BLAS API found.12:37
anishshah-- A library with LAPACK API found.12:37
anishshah-- Could NOT find ATLAS (missing:  ATLAS_LIBRARIES ATLAS_INCLUDES)12:37
anishshah-- Found CBLAS: /usr/lib/libcblas.so.312:37
anishshah-- Looking for cblas_dsyrk in /usr/lib/libcblas.so.312:37
anishshah-- Looking for cblas_dsyrk in /usr/lib/libcblas.so.3 - found12:37
anishshah-- Could NOT find GLPK (missing:  GLPK_LIBRARY GLPK_INCLUDE_DIR GLPK_PROPER_VERSION_FOUND)12:37
anishshah-- Could NOT find CPLEX (missing:  CPLEX_LIBRARY CPLEX_INCLUDE_DIR)12:37
anishshah-- Could NOT find Eigen3 (missing:  EIGEN_INCLUDE_DIR) (Required is at least version "3.1.2")12:37
anishshah-- Could NOT find OPENCL (missing:  OPENCL_LIBRARY OPENCL_INCLUDE_DIR)12:37
anishshah-- Could NOT find ViennaCL (missing:  VIENNACL_INCLUDE_DIR OPENCL_INCLUDE_DIRS OPENCL_LIBRARIES) (Required is at least version "1.5.0")12:37
anishshah-- checking for one of the modules 'libColPack>=1.0.9;ColPack>=1.0.9'12:37
anishshah-- Could NOT find NLOPT (missing:  NLOPT_LIBRARY NLOPT_INCLUDE_DIR)12:38
anishshah-- Could NOT find LPSOLVE (missing:  LPSOLVE_LIBRARIES LPSOLVE_INCLUDE_DIR)12:38
anishshah-- Could NOT find ColPack (missing:  COLPACK_LIBRARIES COLPACK_LIBRARY_DIR COLPACK_INCLUDE_DIR)12:38
anishshah-- Could NOT find ARPREC (missing:  ARPREC_LIBRARIES ARPREC_INCLUDE_DIR)12:38
anishshah-- Could NOT find Doxygen (missing:  DOXYGEN_EXECUTABLE) (Required is at least version "1.8.6")12:38
anishshah-- checking for one of the modules 'libjson>=0.11;json>=0.11;json-c>=0.11'12:38
anishshah-- Could NOT find HDF5 (missing:  HDF5_LIBRARIES HDF5_INCLUDE_DIRS)12:38
anishshah-- Could NOT find CURL (missing:  CURL_LIBRARY CURL_INCLUDE_DIR)12:38
anishshah-- Could NOT find SNAPPY (missing:  SNAPPY_LIBRARIES SNAPPY_INCLUDE_DIR)12:38
anishshah-- Lzo includes and libraries NOT found.12:38
anishshah-- Spinlock support found12:38
anishshah-- Could NOT find PROTOBUF (missing:  PROTOBUF_LIBRARY PROTOBUF_INCLUDE_DIR)12:38
anishshah-- Supported backend(s) for linear algebra NOT FOUND/NOT ENABLED12:38
anishshahAm I missing a lot of dependencies? Is there any easy way to install all of these?12:38
@besser82anishshah, all but cplex and mosek (wihch are commecial), should be provided by your distro12:40
@besser82anishshah, the only required dependency is Eigen312:40
@besser82anishshah, OpenCL / ViennaCL is just of use, when you have OpenCL-enabled hardware12:41
@besser82anishshah, json-c or HDF5 are just needed, if you want to serialize shogun-data from / to that formats12:41
@besser82anishshah, NLopt, LpSolve, ColPack and arprec are optional and not needed usually12:42
@besser82anishshah, except for some special functions12:43
@besser82anishshah, Doxygen is just needed, if you want to build the API-docs12:43
@besser82anishshah, which distro are you using, btw?12:43
anishshahUbuntu 15.1012:44
@HeikoSwiking: around?12:44
anishshahThanks for the detailed info :)12:44
@besser82anishshah, you're welcome ^_^12:44
@HeikoSsonne|work1: around?12:44
@HeikoSbesser82: do you know what this singleton thing means?12:47
@HeikoShttp://www.shogun-toolbox.org/doc/en/latest/singletonshogun_1_1CDenseFeatures.html12:47
@besser82HeikoS, I cannot see a Singleton here  :(12:49
@HeikoSin the link e.g.12:49
@HeikoSbesser82: or in the first line:12:49
@HeikoSCDenseFeatures< ST > Singleton Reference12:49
anishshahbesser82, I'm getting this error while compiling an example. /SGVector.h:69:26: error: expected unqualified-id before ‘using’12:52
anishshahsorry if it's a naive one.12:53
@besser82HeikoS, now I see what's the reason for having "singleton" show up...  It's a non-public template in a cpp-file, which just get's stub-declared in LDA.h  :P12:54
@HeikoSbesser82: ah man12:54
@HeikoShow to fix such things?12:54
@besser82HeikoS, add support for stub-declared templates to Doxygen  :P12:55
@HeikoSargh12:55
@besser82HeikoS, or simply hide'em from Doxygen12:55
@HeikoSthing for me is12:55
@HeikoSwhat to hide?12:56
@HeikoSI just want:12:56
@HeikoSshogun.ml/CDenseFeatures12:56
@HeikoSbesser82: which points to http://shogun.ml/doc/en/latest/classshogun_1_1CDenseFeatures.html12:56
@HeikoSbesser82: which gives 40412:56
@HeikoSsince the pattern does only work for classes, not singletons12:56
@HeikoSwould live to avoid messing around too much, just want my short link to work ;)12:57
@besser82HeikoS, mhh...  Lemme check if there is a workaround...12:58
@besser82anishshah, looks like your compiler needs '-std=c++11' defined12:59
@besser82anishshah, just add it to your compiler-command12:59
@besser82HeikoS, I'll check that later, when I need to dive into Doxygen-stuff anyways13:00
anishshahduring cmake?13:01
@HeikoSbesser82: ok thanks13:01
@HeikoSbesser82: just tried compiling dynaplugz13:01
@HeikoSworked13:01
@HeikoSbut nothing in there yet ;)13:02
@besser82anishshah, you're building the examples manually?13:06
@besser82anishshah, during build of shogun itsels?13:06
@besser82HeikoS, stuff is upcomming, just finishing the last implementation-details  =)13:07
anishshahI did make && make install and it finished successfully13:07
anishshahI'm trying to run shogun/examples/undocumented/libshogun/classifier_knn.cpp13:07
anishshahto check if it is installed correctly.13:07
@besser82anishshah, building it manually?13:08
anishshahI did g++ -std=c++11 classifier_knn.cpp13:08
@besser82anishshah, and what's the outcome?13:08
anishshahBut I'm getting lot's of undefined ref13:08
@HeikoSanishshah: you need to use -lshogun13:08
@HeikoSand the lishogun.so needs to be in your LD_LIBRARY_PATH13:08
@besser82anishshah, yes, because you need to link against against shogun...  Thanks HeikoS ^^13:09
@HeikoSanishshah: standard c++ configuration, give it a google13:09
anishshahOK13:10
anishshahI did "g++ classifier_knn.cpp  -std=c++11 -lshogun", got /tmp/cc8XPUmP.o: In function `main':13:11
anishshahclassifier_knn.cpp:(.text+0x281): undefined reference to `shogun::CSGObject::unref()'13:11
anishshahOK besser8213:11
sonne|work1HeikoS: yes?13:14
@HeikoSsonne|work1: see doxygen problem above13:14
@HeikoSanishshah: shogun is not in the library path then13:14
@HeikoSanishshah: you either need to install it and run ldconfig13:14
@HeikoSor add the path manually via export in bash13:15
sonne|work1HeikoS:  http://www.shogun-toolbox.org/doc/en/latest/singletonshogun_1_1CDenseFeatures.html ?13:15
@HeikoSit is located in the build directory13:15
@HeikoSsonne|work1: yet exactly13:15
@HeikoSshogun.ml/CDenseFeatures is broken because of that singleton thing13:15
aditya123Hi everyone, I am new to the community and willing to contribute. Could anyone guide me through ?13:15
anishshahexport LD_LIBRARY_PATH="/home/anish/Projects/shogun/build/src/shogun:$LD_LIBRARY_PATH"13:16
anishshahI ran this?13:17
anishshahlibshogun.so is in the above directory.13:17
sonne|work1HeikoS: bug in doxygen https://bugzilla.gnome.org/show_bug.cgi?id=73537613:17
@HeikoSaditya123: hi there, please have a look into our wiki, there is a "getting involved" buide13:17
sonne|work1HeikoS: but did get fixed here: https://github.com/doxygen/doxygen/commit/92eb236037e857f38eaf24238815641a4854079213:18
@HeikoSaditya123: best thing is to solve a small entrace task, or send any other small patch13:18
@HeikoSanishshah: try ldconfig then13:18
@HeikoSsonne|work1: checking13:18
@HeikoSsonne|work1: can you install new doxygen on buildbot then?13:19
sonne|work1HeikoS: which version is ours?13:20
sonne|work1that is the newest in debian https://packages.debian.org/sid/doxygen13:20
@HeikoSsonne|work1: I dont know, never touch the buildbot13:20
anishshahHeikoS, sorry for the naive question. I haven't used ldconfig before. Should I run it in build/src/shogun directory?13:21
@HeikoSanishshah: best thing is to google its usage, all you need to do is somehow tell the g++ linker where the shogun library file is13:21
@HeikoSanishshah:13:22
@HeikoShttp://askubuntu.com/questions/631275/how-do-i-do-this-install-you-may-need-to-run-ldconfig13:22
@HeikoSanishshah: and please dont worry about such questions, I have asked them many times myself ;)13:36
anishshahThanks13:40
anishshahTrying13:40
anishshahactually, after "sudo make install", it installed various header file and libshogun.so to "/usr/local/include/" and "/usr/local/lib/libshogun.so"13:41
@HeikoSanishshah: make install sorts out your LD_LIBRARY_PATH  in fact13:41
anishshahI did "cd /usr/local/lib/" and "sudo ldconfig" but getting the same error while compiling13:41
@HeikoSso if you  sudo ldconfig in the main make dir that should do it13:41
anishshahI made a test file which just imports shogun headers and I compiles properly13:42
anishshahmain directory?13:42
anishshahthe one which I "git cloned"?13:42
@HeikoSI meant the "build" dir, but I think it doesnt matter,13:43
@HeikoSanishshah: https://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dynamic_or_shared.html13:45
@HeikoSanishshah: btw where you put the -lsghoun also matters.13:49
@HeikoSI mean before or after the .cpp filename in the g++ command13:50
anishshahhttp://pastebin.ubuntu.com/15258815/ This is the output of my terminal when I put -lshogun before and after13:54
anishshahIs this still a problem because of linking with g++13:54
@HeikoSanishshah: yes, this means that g++ cannot find the shogun dynamic library13:57
anishshahOK13:58
@HeikoSif you are 100% sure that the path is set correctly, move the -lshogun around in the g++ command13:58
@HeikoSbefore or after source file13:58
@HeikoSand also have a read why that might help13:58
anishshahOK. Thanks for your help. I will have a look at this again.14:01
@HeikoSlisitsyn: around?14:13
lisitsynHeikoS: JA14:13
@HeikoSlisitsyn: I think we will have to do wrappers for SGMatrix and SGMatrix seperately14:13
lisitsynwhy?14:13
lisitsynwhat?14:13
@HeikoScant fix the build otherwise14:13
lisitsyn:D14:13
@HeikoSset_generic14:13
lisitsynSGMatrix and SGMatrix?14:13
@HeikoSis all fucked up14:13
@HeikoSonly allows for basic types14:14
@HeikoSlisitsyn: it basically cannot be de-serialised14:14
@HeikoSthe way it is14:14
@HeikoSso we need14:14
@HeikoSSGWrappedBaiscObject14:15
@HeikoSSGWrappedSGMatrixObject14:15
@HeikoSSGWrappedSGVectorObject14:15
@HeikoSso that template parameter only takes basic type for each14:15
@HeikoSit is not possible to do CSGObjectWrapper<SGVector<T>>14:15
@HeikoSas this cannot be instanciated empty14:16
@HeikoSwhich is needed for serial14:16
@HeikoSization14:16
lisitsynHeikoS: damn14:16
@HeikoSat the moment14:16
@HeikoSthats what all the generic stuff is for14:16
@HeikoSso I will add three classes instead of 114:16
@HeikoShaha14:16
lisitsynHeikoS: is there any other workaround?14:18
@HeikoSlisitsyn: I dont think14:18
@HeikoSfix serialization14:19
@HeikoSis that feasible? :D14:19
-!- bollu_ [01ba00d7@gateway/web/freenode/ip.1.186.0.215] has joined #shogun14:24
bollu_hello everyone14:24
bollu_this is about the shogun ML toolkit right?14:24
lisitsynhello14:24
lisitsynyes14:24
bollu_so, I'd like to pick up the "detox shogun" project idea for gsoc 201614:25
bollu_since I've started learning ML (I don't know much, though I have followed through most of the coursera course on ML by Andrew ng. I need to read the proper math behind much of it).14:25
bollu_but I have worked on pretty large codebases in different languages14:26
bollu_here's my github for reference: http://github.com/bollu14:26
bollu_so, well, I wanted to know where I can start14:26
bollu_re-writing code to fit C++11 while picking up ML seems like an awesome way to spend the summer :)14:26
lisitsynbollu_: I'd recommend you either to check available issues14:26
lisitsynor do some research on code14:26
lisitsynand improve random parts :)14:27
bollu_lisitsyn: heh, cool. That seems like a good way to start as any other14:27
bollu_lisitsyn: how likely is it that this proposal would get selected (provided I do my part, ofc - sending pull requests and fixing code)14:27
bollu_I was thinking of using "cereal" for serialization14:28
bollu_since the project idea says that serialization is a major pain point14:28
bollu_does that sound like a decent place to start?14:28
lisitsynbollu_: yes for sure14:28
-!- ebadawy [~ebadawy@197.163.32.61] has quit [Ping timeout: 252 seconds]14:36
@HeikoSbollu_: serialization is a major challenge14:52
@HeikoSbollu_: better start with somethin simpler14:53
@HeikoSbollu_: if you wanna do the detox project, best to improve any existing shogun algorithm14:53
@HeikoSyou could start with fixing the knn documentation, there is an issue on that14:53
@HeikoSthen benchmark it, like saurabh did for kmeans14:53
@HeikoScerialization only if you did a few smaller things, as it is tricky14:54
bollu_HeikoS: can I have a link?14:54
bollu_to the github issue?14:54
@HeikoSuse the search, I dont have it in my head :)14:54
bollu_xD14:58
bollu_alright14:58
@HeikoSlisitsyn:15:17
@HeikoSI fixed it15:17
@HeikoSlisitsyn: can you fix the notetests of genertor?15:20
@HeikoSI think store-vars is on by default15:20
@HeikoSthe test passed before you added it15:20
@HeikoSlisitsyn: also appreciate feedback on the fix: https://github.com/shogun-toolbox/shogun/commit/8bbf4de6bfa8c4c9f11c67ac8a63db764caecd0215:21
-!- HeikoS [~heiko@host-92-0-162-192.as43234.net] has quit [Quit: Leaving.]15:22
-!- anishshah [~anishshah@27.5.145.62] has quit [Ping timeout: 248 seconds]15:28
-!- bollu_ [01ba00d7@gateway/web/freenode/ip.1.186.0.215] has quit [Ping timeout: 252 seconds]15:42
-!- bollu [~Adium@1.186.0.215] has joined #shogun15:49
-!- bollu1 [~Adium@1.186.0.105] has joined #shogun15:58
-!- bollu [~Adium@1.186.0.215] has quit [Ping timeout: 252 seconds]15:59
-!- bollu1 [~Adium@1.186.0.105] has quit [Quit: Leaving.]16:04
-!- anishshah [~anishshah@27.5.145.62] has joined #shogun16:26
-!- lambday [6a33e982@gateway/web/freenode/ip.106.51.233.130] has joined #shogun16:47
-!- mode/#shogun [+o lambday] by ChanServ16:47
-!- anishshah [~anishshah@27.5.145.62] has quit [Ping timeout: 260 seconds]16:48
-!- anishshah [~anishshah@27.5.145.62] has joined #shogun17:06
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun17:14
-!- mode/#shogun [+o HeikoS] by ChanServ17:14
@lambdayHeikoS: hey17:36
-!- lambday [6a33e982@gateway/web/freenode/ip.106.51.233.130] has quit []17:48
-!- lambday [6a33e982@gateway/web/freenode/ip.106.51.233.130] has joined #shogun17:53
-!- mode/#shogun [+o lambday] by ChanServ17:53
-!- anishshah [~anishshah@27.5.145.62] has quit [Ping timeout: 252 seconds]18:39
-!- akshat_jain [quassel@nat/iiit/x-jecpppwhpgpefkzh] has joined #shogun18:42
-!- akshat_jain [quassel@nat/iiit/x-jecpppwhpgpefkzh] has quit [Client Quit]18:46
-!- akshat_jain [quassel@nat/iiit/x-rgionbjjqnuuiyji] has joined #shogun18:46
-!- akshat_jain [quassel@nat/iiit/x-rgionbjjqnuuiyji] has quit [Remote host closed the connection]18:57
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Ping timeout: 250 seconds]18:57
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun19:01
-!- mode/#shogun [+o HeikoS] by ChanServ19:01
@HeikoSlambday: jojojo19:01
@HeikoSyeah 5 admins in irc :)19:02
@lambdayHeikoS: yoo19:02
@lambday:D19:02
@lambdayHeikoS: I was just writing the mail to you and arthur19:02
@HeikoSah cool19:02
@HeikoSlambday: I think its good to say hello and have a brief (very brief) chat every day or too19:03
@HeikoSjust to keep things moving19:03
@lambdayHeikoS: yeah.. also I'm adding a few things that I thought I'd discuss19:03
@lambdayjust a brief : I noticed that we don't actually have linear time HSIC implemented in shogun19:04
@lambdayHeikoS: remember the refactoring idea we discussed about, so that a streaming hsic fits nicely19:04
@lambdayHeikoS: so I was thinking of spending some time this week on that.. we're planning to add some more estimates in future as well.. so I'd try to make sure I get the design correct19:05
@lambdayHeikoS: and then next thing is the toy examples19:06
@HeikoSlambday: yes agreed, linear hsic is good19:06
@lambdayHeikoS: many of them are already in python modular - just gotta make sure these things are working19:06
@HeikoSlambday: toy examples as well19:06
@lambdayHeikoS: also, I was (just blindly) trying to fix the mmd notebook http://nbviewer.jupyter.org/gist/lambday/e064dc658f1b20755f2a19:07
@lambdayso these 3 things for week 119:07
@lambdayand then independence test notebook..19:07
@HeikoSgood that the notebook is fixed19:07
@HeikoSyes, I agree that sounds like a good plan19:08
@lambdayHeikoS: hehe I am not yet sure.. gotta make sure that the numbers make sense :D19:08
@HeikoSthe examples could even go to new meta lanugage19:08
@HeikoSI think all the plots look nice19:08
@HeikoSlambday: it would be cool if we could get the linear time mmd working from a giant file19:09
@HeikoSif we had an example in the notebook that would be cool19:09
@HeikoSas well as the toy examples19:09
@lambdayHeikoS: yeah..19:09
@HeikoSlambday: because these gaussian blobs are synthetic19:10
@HeikoSlambday: and btw do you know whether the streaming features interface is thread safe?19:10
@HeikoSbecause I guess we could multicore the statistic19:10
@HeikoSbut maybe later19:10
@HeikoSnot too many things at once19:10
@lambdayHeikoS: I haven't worked on that yet..19:10
@lambdayyeah19:11
@HeikoSlets start simple19:11
@HeikoSworking notebook is a good first step19:11
@HeikoStoy API meta examples are next19:11
@HeikoSand the streaming hsic19:11
@HeikoSthats already a lot19:11
@HeikoSwhat is a good time to IRC idling for you?19:11
@lambdayHeikoS: also, if I do in fact refactor, then the API might change a bit.. in that case, we'll need to have some working examples ready19:12
@lambdayHeikoS: I am planning to be here from 5PM London time onwards (till as long as I'm awake :D)19:12
@HeikoSlambday: you can decide what you want to do first I guess19:13
@HeikoSlambday: ok 5pm london time is perfect for me19:14
@HeikoSlike 5-7 are good hours to catch up here19:14
@lambdayHeikoS: BTW this was the discussion forum for refactoring - https://github.com/shogun-toolbox/shogun/issues/249519:14
@lambdayHeikoS: cool then!19:14
@lambdayHeikoS: check the diagram I had.. I started coding on that in a poc lib..19:15
@HeikoSyes I remember19:15
@HeikoSlambday: this is a huge set of changes19:15
@lambdayI thought of spending first few days on this - cause this may change the API a bit (not too much)... so examples/notebooks have to be modified accordingly19:15
@HeikoSyes good idea19:15
@lambdayHeikoS: I already have (parts of a) poc ready19:16
@lambdayhttps://github.com/lambday/flash19:16
@HeikoSI remember that thing19:16
@lambdayit separates data fetching/streaming and kernel part completely..19:16
@HeikoSyes that is good19:17
@HeikoSvery good19:17
@HeikoSbtw19:17
@lambdayso same streamer can be used for linear time hsic and linear time mmd19:17
@lambdaysimilar things for blocks19:17
@HeikoSwe might want to implement 3 variable or 1 variable (goodness of fit) tests too at some point19:17
@HeikoSyes very nice19:17
@HeikoSlambday: btw we have this some<T> thing by lisitsyn now19:18
@lambdayHeikoS: yeah.. our wrapper for std::unique_ptr<T> IIRC19:18
@lambday:D19:18
@HeikoSexactly ;)19:18
@lambdayor shared :/19:18
lisitsynbetter name19:18
lisitsynshared19:18
@HeikoSlisitsyn: better change now19:19
@HeikoSbefore it is all over the place ;)19:19
@HeikoSlisitsyn: saw my fix btw?19:19
lisitsynsome is a good name19:19
lisitsynHeikoS: not yet sorry19:19
@HeikoSlisitsyn: but now XML serialisation is broken for CHAR19:19
@HeikoSWTF!!!19:19
@lambdaylisitsyn: long time man...19:19
lisitsynlambday: h e l l o19:19
@HeikoShaha whats the last time you guys talked?19:20
@HeikoSlisitsyn: make sure to come visit me in London when lambday is here in summer!19:20
@lambdayHeikoS: we talked about redesigning CSGObject19:20
@lambday:D19:20
@HeikoSmeet in person19:20
@HeikoSlambday: I see ;D19:20
@HeikoShaha19:20
lisitsynHeikoS: adjust the oil price19:20
@HeikoSI was sleeping probably19:20
@lambdaythat was back in march/april I see19:20
@HeikoSlisitsyn: earn in GPBs !19:20
lisitsynHeikoS: send me your money19:20
@HeikoSlisitsyn: what money?19:21
lisitsynall money19:21
@HeikoSi am student forgot?19:21
lisitsynthat's how I earn in GPB19:21
lisitsynthe money19:21
@HeikoShehe19:21
lisitsyngive me19:21
lisitsyngive it to me19:21
@HeikoSIll give it to you when you come visit ;=)19:21
@lambdayhahaha19:21
@HeikoSbut then I charge it as rent ;=)19:21
lisitsynhello sir I have ONE BILLION DOLLAR on my blocked account19:22
lisitsynplease send me ONE THOUSAND DOLLAR so I can unblock and send it to you19:22
@HeikoShello sir ;)19:22
@HeikoShttps://travis-ci.org/shogun-toolbox/shogun/jobs/112864762#L457719:23
@HeikoSplease guide me accordingly in fixing xml serialization19:23
-!- anishshah [~anishshah@27.5.145.62] has joined #shogun19:24
lisitsynHeikoS: but first you send me the money19:25
lisitsynUnnamed?19:25
lisitsynwrappedbasic19:25
lisitsynemha19:25
@HeikoSits only CHAR19:26
@HeikoSwhich means XML cannot serialize the default value I chose in standard constructor, which is 019:26
@HeikoSmaybe a cast helps19:26
lisitsynwoah19:26
@HeikoShow to even turn on xml serialization unit tests?19:27
lisitsynno idea19:27
@HeikoSlisitsyn: would be really good to have a reliable serialization19:27
@HeikoSthat doesnt restrict us in doing simple things19:27
@HeikoSlike wrapping elements19:27
@HeikoSthat are non-basic types gnaaa19:27
lisitsynHeikoS: that's something hackable in a few evenings19:29
lisitsynand yes I still didn't do that haha19:29
@HeikoSlisitsyn: haha19:29
lisitsynbecause I am dying!19:29
lisitsynhaha19:29
@HeikoSso I spend multiple evening trying to hack around it19:29
@HeikoSI fixed it19:30
@HeikoShahaha19:30
@HeikoScheck it out19:30
@HeikoSm_value = 'a';19:30
@HeikoSwhere m_value is of type T19:31
@HeikoSthen XML passes19:31
lisitsynmeh19:31
@HeikoSso it cannot serialize the char for 019:31
lisitsynyes19:31
lisitsynbecause it is like end of string19:31
lisitsyngosh19:31
lisitsyn:)19:31
lisitsynit is not the best char ever19:31
@HeikoSwhats a good default value? :)19:31
@HeikoSI think i put 'H'19:32
@HeikoSfor HEIKO19:32
@HeikoSor maybe random?19:32
@HeikoSah not that breaks serialization again19:32
lisitsynHEIKO is good19:33
lisitsynhaha19:33
@HeikoSproblem is that if I use 'a' then half of float32_t is uninitialised19:33
lisitsynmake your own trace in the code19:33
@HeikoSlisitsyn: how to fix?19:33
lisitsyndon't know19:33
@HeikoSthough test seems to pass19:33
@HeikoSweirdly19:34
@HeikoSmaybe g19:34
@HeikoSg++ is smart and puts 0 to the rest?19:34
@HeikoSPython:19:35
@HeikoSif type(T) is char:    collect_garbage()19:36
@HeikoS;19:36
@HeikoStell_wiking19:36
-!- Guest92392 [ibm___@nat/iiit/x-ezdaseupxeljwuxn] has quit [Remote host closed the connection]19:37
@HeikoSlisitsyn:  https://github.com/shogun-toolbox/shogun/commit/798a3ac6d912315e894afb3168579008da92acf019:37
-!- ebadawy [~ebadawy@197.163.32.61] has joined #shogun19:38
-!- ibm [ibm___@nat/iiit/x-jvlytqzdmtlyqjyw] has joined #shogun19:38
-!- ibm is now known as Guest4203719:38
-!- nelson-liu [~nelsonliu@nelsons-macbook-pro.dyn.cs.washington.edu] has joined #shogun19:39
lisitsynHeikoS: ok you need better solution19:41
lisitsynHeikoS: template specialization19:42
@HeikoSlisitsyn: ah I knew you were gonna say that19:42
-!- anishshah [~anishshah@27.5.145.62] has quit [Ping timeout: 276 seconds]19:45
-!- ratul [0e8becd2@gateway/web/freenode/ip.14.139.236.210] has joined #shogun19:46
@HeikoSlisitsyn:19:47
@HeikoScan you turn off appending things in the generator by default?19:47
@HeikoSso that the tests still pass?19:47
@HeikoSwithout this19:47
@HeikoS-19:47
@HeikoS- __sg_storage <- WrappedObjectArray()19:47
@HeikoS- __sg_storage_file <- SerializableAsciiFile("None.txt", 119)19:47
@HeikoS- __sg_storage$save_serializable(__sg_storage_file)19:47
@HeikoSbecause if we have this, then tests are green again19:48
lisitsynHeikoS: yes in a few minutes19:48
@HeikoSlisitsyn: and maybe the template thing as well, you are faster in that19:48
-!- anishshah [~anishshah@27.5.145.62] has joined #shogun19:48
-!- anishshah [~anishshah@27.5.145.62] has quit [Client Quit]19:49
@lambdayyou guys are writing all unit-tests in meta?19:52
@lambdayor only the autogenerated ones?19:52
@HeikoSlambday: not unit tests19:52
@HeikoSlambday: but integration19:52
@HeikoSand only numerical data19:52
@HeikoSnot classes themselves19:52
@lambdaycool!!!19:52
@HeikoSto avoid all the false positives19:53
@HeikoSyeah it's cool ,we can then compare across all interfaces with a single example19:53
@HeikoSas well as documenting it on the fly (hiding the testing details in the cookbook page)19:53
@lambdayare these things available yet on develop?19:53
@HeikoSlambday: nope unfortunately not19:54
@HeikoSlambday: but we are close19:54
@HeikoSlambday: very close to merging I would say19:54
@HeikoSlambday: but we tend to break the build often so keeping things in a seperate branch for now19:55
@HeikoSlambday: saw the screenshot of the cookbook?19:55
@HeikoSlambday: https://cloud.githubusercontent.com/assets/3594351/13373164/c1399ca6-dd57-11e5-866b-4db9f3f2fb0f.png19:55
@lambdayHeikoS: the one that you shared on hangout that day?19:55
@HeikoSall languages are in19:55
@HeikoSlambday: ah yes I showed you19:56
@lambdayyeah this looks epic19:56
@HeikoSI added all bindings19:56
@HeikoSapart from lua, which is broken19:56
@HeikoSlambday: oh and we also have cpp now19:56
@lambdayI *have* to pick up on the meta language asap! It sounds like magic!19:56
@HeikoSwhich is the best19:56
@HeikoSsince we can compare output that has not gone through the swig typemaps19:57
@HeikoSand this means, we implicitly test all the typemaps to be correct19:57
@HeikoS(which currently they are not=19:57
-!- ratul [0e8becd2@gateway/web/freenode/ip.14.139.236.210] has quit [Ping timeout: 252 seconds]19:58
@HeikoSlambday:  btw https://github.com/shogun-toolbox/shogun/pull/2928/files#diff-c02a2e0b28912409b0450601383094e4R28519:59
@HeikoSwhy only NATIVE here?19:59
@HeikoSlambday: another btw. We will from now on assume that Eigen3 is always available20:00
@HeikoSlambday: so we can get rid of all HAVE_EIGEN3 ifdefs20:00
@HeikoSas well as the native backend kind of can go, and we just use eigen3 if nothing else is available20:00
@HeikoSlambday: what do you think=?20:00
@lambdayHeikoS: that's great! native is such a pain20:01
@HeikoSlambday: I agree20:01
@HeikoSlets just get rid of it20:01
@lambdayHeikoS: I don't remember.. maybe it was to do with restricting compilation failure in absense of eigen320:01
@HeikoSlambday: yes, but that doesnt make sense20:01
@lambdayso from now on, shogun won't compile without eigen3, right?20:01
@HeikoSsince we can bundle it20:01
@HeikoSwe were afraid back then20:02
@HeikoSof dependencies20:02
@HeikoSlambday: I think it does, some tests fail but that is all20:02
@HeikoSif you have a minute, could you tell this guy here: https://github.com/shogun-toolbox/shogun/pull/292820:02
@HeikoSI dont really like this: <linalg::Backend::NATIVE>20:03
@HeikoSso long20:03
@HeikoSwhat about just scale(bla) ?20:03
@lambdayHeikoS: that's supposed to work (with whatever default backend is there)..20:03
@lambdayHeikoS: I'll check.. I really don't remember what the issue was20:04
@HeikoSlambday: but why does the guy put this in all shogun algorithms?20:04
@HeikoSdont we want to be *not* explicit most of the time?20:04
-!- amod95 [uid67143@gateway/web/irccloud.com/x-kyxelqyfqaotbqbu] has joined #shogun20:07
-!- edgex [549f16eb@gateway/web/freenode/ip.84.159.22.235] has joined #shogun20:07
@lambdayHeikoS: yes but I guess there was an issue in viennacl build..20:08
@lambdayHeikoS: yes not being explicit is the goal...20:08
-!- nelson-liu [~nelsonliu@nelsons-macbook-pro.dyn.cs.washington.edu] has quit [Remote host closed the connection]20:09
@lambdayHeikoS: plus if we always assume that eigen3 is there, then we can totally get rid of this native thing20:09
@HeikoSlambday: ok then20:09
@lambdayHeikoS: also, I seriously think that globally setting the backend to viennacl *has* to be disabled20:09
@lambdayone should not be able to do that20:09
@HeikoSlambday: maybe put that as a point to start with as well20:09
@HeikoSlambday: and I agree on that20:09
@lambdaythis calls for weird bugs that are hell to track down20:10
@HeikoSmaybe spend 30-50% on your time on such linalg cleanups for now. It is imporant to polish these things before GSoC20:10
@lambdayHeikoS: yeah I'll make some changes in there as well20:10
@HeikoSyeah definitely20:10
@lambdayHeikoS: so, default backend eigen3, native may stay, but only when someone explicitly wants to set this (for example, he wrote his own linalg code for shogun)..20:11
-!- edgex [549f16eb@gateway/web/freenode/ip.84.159.22.235] has quit [Client Quit]20:11
@lambdayHeikoS: for viennacl/opencl, one MUST explicitly specify20:11
@lambdayor specify per class/per module20:11
@lambdaywill think a bit about that20:12
@HeikoSyes20:12
@HeikoSabsolutely20:12
@HeikoSas GPU programming should be explicit20:12
@HeikoSbut backend is still changable20:12
@HeikoSwhich is good20:12
@HeikoSso we are general while explicit20:12
@HeikoSlambday: I would just delete native to make things easier20:13
@HeikoSit is a shame for the code in there, but this way we also clean up the files and make them less scary to newcomers20:13
@lambdayHeikoS: most of the things don't even have native20:13
@HeikoSyeah20:13
@lambdayyeah true20:13
@lambdayHeikoS: how are we gonna make sure that eigen3 is there?20:14
lisitsynHeikoS: https://github.com/shogun-toolbox/shogun/commit/ead7bfaf8fa27653d95676bd79b594b0647bf36d20:16
@lambdayHeikoS: for dev-testing purpose, I am thinking of having a switch though  - SGMatrix with eigen3 | GPUMatrix with viennacl - a dev can set both from a switch for an entire algorithm.. then compare performance20:16
@lambdayno need to set viennacl shogun-wise20:17
@lambdayotherwise all this data copying back and forth makes it too hard20:17
amod95Hello everyone!20:19
amod95Sorry to interrupt.20:19
@lambdayamod95: hey there!20:19
@HeikoSlisitsyn: thanks!20:19
amod95I am trying to build shogun on Mac OS X 10.11.20:20
@HeikoSlambday: we should modify the build so that eigen3 is *always* bundled. lisitsyn do you agree? wiking?20:20
amod95While building repo from source, I am getting this error: http://pastebin.com/KmKt0N3B20:20
@lambdaylisitsyn: do we have any wrapper for std::unique_ptr yet?20:20
@HeikoSlambday: very good idea with the switch20:20
@lambdayHeikoS: wiking has to comment about the bundle thing20:20
amod95log when I run command 'cmake ..' : http://pastebin.com/jXFfbjv920:21
amod95can someone help?20:22
@HeikoSamod95: weird error. could you open an issue on github with that?  Then someone will help you there, it is more visible than here20:22
amod95okay, I am doing it.20:22
-!- alaaelnouby [~alaaelnou@41.47.4.176] has joined #shogun20:22
lisitsynlambday: shogun/lib/unique20:30
lisitsynI think20:30
lisitsynI added it for pimpl20:30
@lambdaylisitsyn: superb!20:30
@HeikoSlambday, we already talked about it and he agreed20:38
@HeikoSso green lights20:38
@lambdayHeikoS: cool then.. I'll redesign linalg a bit accordingly20:39
@HeikoSlambday: super cool20:39
@HeikoSlambday: if you have more time (I know I ask a lot), a linear solve would be great :)20:39
@lambdayHeikoS: maybe opening an issue would help me to keep things on track20:39
@HeikoSthats top on the list of needed operations20:39
@HeikoSlambday: feel free to20:39
@lambdayHeikoS: hehe yeah that's pending for quite a long time20:40
@HeikoSlambday: cool stuff.20:40
lisitsynHeikoS: how to run tests for this stuff20:42
-!- travis-ci [~travis-ci@ec2-54-161-110-71.compute-1.amazonaws.com] has joined #shogun20:42
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/11292741120:42
-!- travis-ci [~travis-ci@ec2-54-161-110-71.compute-1.amazonaws.com] has left #shogun []20:42
@lambdayHeikoS: so I'll do linalg stuffs in a separate branch assuming cmake always finds eigen3.. when the required cmake changes are there in develop for that, we can merge it20:42
@HeikoSlambday: ok I will do that20:43
-!- Saurabh7 [~Saurabh7@103.225.100.51] has joined #shogun21:09
@HeikoSlambday: what about these HAVE_LINALG_LIB switches?21:16
@lambdayHeikoS: IIRC this checks whether (1) we have (any of the) supported backends AND (2) the compiler supports c++1121:17
@HeikoSlambday: I think we might want to assume this soon too21:17
@HeikoSc++121:17
@lambdayHeikoS: if we (1) bundle eigen3 and (2) make c++11 mandatory for shogun, it can go too21:17
@HeikoSok21:17
@HeikoSwiking: ^ you were against making eigen3 mandatory right?21:18
@lambdaysonney2k was against it :P21:18
-!- bollu [~Adium@1.186.131.55] has joined #shogun21:18
@lambdayabout c++11, we more or less agreed : https://github.com/shogun-toolbox/shogun/issues/271921:19
-!- alaaelnouby [~alaaelnou@41.47.4.176] has quit [Quit: Leaving]21:23
@HeikoSlambday: yeah21:28
@HeikoSshould do that after the eigen3 refactor21:28
@HeikoSlambday: check out this21:28
@HeikoShttps://github.com/shogun-toolbox/shogun/tree/feature/eigen3_bundling21:28
@HeikoSlisitsyn, wiking ^21:28
@HeikoSwaiting what travis sais21:30
@lambdayHeikoS: some of the SG_ERROR/SERROR messages are still there I think21:34
@lambdaySG_ERROR("Only possible if shogun is compiled with EIGEN3 enabled\n");21:34
@HeikoSlambday: I think i fixed all21:34
@HeikoScompile whole libshogun21:34
@HeikoSand this cannot happen if there is an ifdef that is forgotten21:35
@HeikoSah one case21:35
@lambdaybetween #else and #endif21:36
@lambdayyeah21:36
lisitsynHeikoS: what's broken now21:36
lisitsynin sphinx_doc21:36
@HeikoSlisitsyn: run nosetests in generator dir21:37
@HeikoSlisitsyn: unit tests for input output of meta examples21:37
@HeikoSsince you added the storing stuff, they break21:37
@HeikoSshould be off by default, i.e. behave 100% as before21:37
lisitsynah ok21:37
@HeikoSlambday: fixed them!21:38
@HeikoSlisitsyn: lambday going home now21:38
@HeikoSgood night!21:38
@lambdayHeikoS: good night man!21:39
@lambdayHeikoS: after a long time I'm awake till 2am for good cause :D21:39
lisitsynHeikoS: done21:41
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Ping timeout: 248 seconds]21:42
-!- D3M0T3P [~D3M0T3P@171.53.77.83.dynamic.wline.res.cust.swisscom.ch] has joined #shogun21:48
-!- D3M0T3P [~D3M0T3P@171.53.77.83.dynamic.wline.res.cust.swisscom.ch] has left #shogun ["Leaving"]21:55
Saurabh7hey lambday22:03
-!- kpun [0e8b5206@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.82.6] has joined #shogun22:27
@sonney2klambday, hey long time no see!23:08
@sonney2kwiking, around?23:12
-!- shubham__ [uid57367@gateway/web/irccloud.com/x-erbfovrtgtdrllcz] has joined #shogun23:15
-!- travis-ci [~travis-ci@ec2-54-145-30-142.compute-1.amazonaws.com] has joined #shogun23:30
travis-ciit's Sergey Lisitsyn'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/11293876323:30
-!- travis-ci [~travis-ci@ec2-54-145-30-142.compute-1.amazonaws.com] has left #shogun []23:30
-!- kpun [0e8b5206@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.82.6] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]23:36
@sonney2kwiking, how difficult would it be to have docker images for buildbot based on say debian testing? latest ubuntu etc?23:52
@sonney2kwiking, so we can separate the OS on the server from the stuff run inside the buildbot?23:52
-!- HeikoS [~heiko@host-92-0-162-192.as43234.net] has joined #shogun23:58
-!- mode/#shogun [+o HeikoS] by ChanServ23:58
-!- amod95 [uid67143@gateway/web/irccloud.com/x-kyxelqyfqaotbqbu] has quit [Quit: Connection closed for inactivity]23:58
--- Log closed Wed Mar 02 00:00:19 2016
« Monday, 2016-02-29 Index Wednesday, 2016-03-02 »

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