Open in new window / Try shogun cloud
--- Log opened Tue Sep 17 00:00:02 2013
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 245 seconds]00:02
van51sonney2k: I made a PR for the RKS..00:22
van51sonney2k: I don't think it's trivial to allow subclasses to implement directly their functions in the dot methods if they want00:25
van51sonney2k: through a supplementary method I mean00:25
van51sonney2k: but I guess they can just override the methods as whole :)00:26
van51sonney2k: if you have another suggestion let me know00:26
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has quit [Quit: Leaving.]00:26
-!- zxtx [~zv@129-79-241-50.dhcp-bl.indiana.edu] has quit [Ping timeout: 245 seconds]00:50
-!- zxtx [~zv@ip-64-134-197-230.public.wayport.net] has joined #shogun02:35
shogun-buildbotbuild #526 of nightly_default is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/52603:19
-!- foulwall [~zhengyang@114.255.40.22] has joined #shogun07:04
-!- sonne|osx [~sonne@89.204.138.216] has joined #shogun08:21
-!- lisitsyn [~lisitsyn@fb2-lo1.global63.net] has joined #shogun08:28
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun08:29
shogun-notifier-shogun: Soeren Sonnenburg :develop * 082873b / data: https://github.com/shogun-toolbox/shogun/commit/082873b0e6b6a5581c90bb4717c0292db4fe3b5108:29
shogun-notifier-shogun: require new data version for hdf5 test08:29
sonne|osxwiking: in case you are awake / have time - looks like the buildbots filled up the harddisk08:31
sonne|osx(2TB ...)08:31
sonne|osxso please erase a bit08:31
shogun-buildbotbuild #2166 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2166  blamelist: Soeren Sonnenburg <sonne@debian.org>08:32
-!- sonne|osx [~sonne@89.204.138.216] has quit [Quit: sonne|osx]08:35
wikingsonney2k: ?09:08
wikinghow the fuck is this possible09:09
wikingno not really09:09
wiking/dev/sda2       1.8T  110G  1.6T   7% /home09:09
-!- lisitsyn [~lisitsyn@fb2-lo1.global63.net] has quit [Quit: Leaving.]09:23
-!- besser82 [~besser82@fedora/besser82] has joined #shogun09:23
besser82Mornin' Folks!09:24
besser82wiking: ARPREC is avail now in Fedora ;)09:25
besser82wiking: yum install arprec-devel09:25
wikinghey :)09:26
besser82wiking: and the new update of rubygem-narray should fixup the issues with ruby_modular ;)09:33
besser82wiking: but there's an issue with R_modular: make install fails, if R_modular is enabled && build :(09:34
wikingwhat is it09:35
wikingi'm just about to send a patch for your prvious problem09:35
-!- travis-ci [~travis-ci@ec2-54-242-16-107.compute-1.amazonaws.com] has joined #shogun09:36
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1145154609:36
-!- travis-ci [~travis-ci@ec2-54-242-16-107.compute-1.amazonaws.com] has left #shogun []09:36
shogun-notifier-shogun: Viktor Gal :develop * 0d0f3ee / tests/unit/base/clone_unittest.cc.py: https://github.com/shogun-toolbox/shogun/commit/0d0f3eecfec68c0c09e8185f370d9092c4b4f25309:40
shogun-notifier-shogun: Fix dummy test case generation for automated unittest if jinja2 is not09:40
shogun-notifier-shogun: present09:40
besser82wiking: just second ;) need to rebuild with it enabled ;)09:40
shogun-buildbotbuild #2167 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2167  blamelist: Viktor Gal <viktor.gal@maeth.com>09:42
wikingbesser82: here09:42
besser82wiking: currently building...  will paste output of failure ;)09:45
wikingok09:52
sonne|workwiking: check the buildbot's master09:57
sonne|workman why are we suddenly having http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/216709:58
wikingsonne|work: i guess one commit did this09:59
wikingsonne|work: /dev/sda2       914G   37G  831G   5% /home09:59
sonne|workweird then09:59
besser82wiking: -- Installing: /home/besser82/rpmbuild/BUILDROOT/shogun-3.0.0-0.1.git20130915.f56e63e.fc20.x86_64/usr/lib64/R/library/shogun/R/modshogun.RData10:02
besser82CMake Error at src/interfaces/r_modular/cmake_install.cmake:60 (FILE):10:02
besser82  file problem creating directory: /usr/lib64/R/library/shogun/Meta10:02
besser82Call Stack (most recent call first):10:02
besser82  cmake_install.cmake:41 (INCLUDE)10:02
besser82wiking: ^^^10:02
wikingdo u have /usr/lib64/R/library/shogun ?10:03
besser82wiking: so you say the dir needs to be present BEFORE install?10:04
wikingis it present now?10:04
besser82wiking: yes10:04
besser82wiking: %{buildroot}/usr/lib64/R/library/shogun  is present10:05
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun10:05
wikingok10:05
wikingthen why cannot it create a Meta dir in it?10:05
wiking:)10:05
besser82wiking: I dunno?!?10:05
besser82wiking: enough free space on disk > 600GBytes10:05
wikingmmm interesting10:05
besser82wiking: btw. Perl_modular doesn't even finsh configure stage...10:07
wikingbesser82: because that's not even enabled...10:07
wikingthat's an untested modular interface10:07
wikingso dont even try to put into a package10:07
besser82wiking: kk :)10:07
wikingsonne|work: interestingly enough the same repo compiles on my machine just fine..10:08
shogun-notifier-shogun-data: lambday :master * 48d5dec / logdet/west0479.mtx.gz,logdet/west0479.txt.gz,logdet/west0479.tar.gz: https://github.com/shogun-toolbox/shogun-data/commit/48d5dec2cdfde862cf22b64122f6fcf541691cba10:43
shogun-notifier-shogun-data: Added matrix-data in mtx (original) and ascii (converted) (log-det)10:43
shogun-notifier-shogun-data: Heiko Strathmann :master * 0bf9160 / logdet/west0479.mtx.gz,logdet/west0479.txt.gz,logdet/west0479.tar.gz: https://github.com/shogun-toolbox/shogun-data/commit/0bf91608b044630a85f2f637086624a2c04d605210:43
shogun-notifier-shogun-data: Merge pull request #20 from lambday/master10:43
shogun-notifier-shogun-data:10:43
shogun-notifier-shogun-data: Added matrix-data in mtx (original) and ascii (converted) (log-det)10:43
-!- lambday [67157d37@gateway/web/freenode/ip.103.21.125.55] has joined #shogun10:46
wikinglisitsyn: about the whole thing apply/train10:52
wikinglisitsyn: what if we start mimicing gstreamer setup?10:52
lisitsynwiking: I don't know about gstreamer10:52
wikinglisitsyn: well it's a bunch of .so -s10:52
wikingand then you just hook together containers10:52
wikinginto a pipeline10:53
lisitsynyeah it seems like it is the way10:53
wikinghence when u run the pipeline10:53
wikingthe only thing that is loaded into the memory is the required .sos10:53
wikingi mean dylibs10:53
lisitsynyeah or dlls10:54
lisitsynwhatever :)10:54
wikingyeah ...10:54
-!- Guru___ [c2c71719@gateway/web/freenode/ip.194.199.23.25] has joined #shogun10:54
wikingthis way we save  up as well a lot of memory :P10:54
lisitsynI find it quite bad that we have like >200mb so/dylib/dll10:54
wikingyeah10:54
wikingi mean gstreamer would be the same10:54
wikingwith all those video codecs it suppors10:54
wiking*supports10:54
-!- Guru___ [c2c71719@gateway/web/freenode/ip.194.199.23.25] has left #shogun []10:55
wikinghehe not really true but whatever:10:55
wiking 21M/opt/local/lib/gstreamer-1.0/10:55
wikingit would be bigger than it is now10:55
-!- thoralf [~thoralf@enki.zib.de] has joined #shogun11:03
thoralfHey.11:03
besser82wiking: Why does `make install` place libshogun.so* and other things in /usr/lib although they should go to /usr/lib64 on x86_64?11:04
besser82wiking: python_modular and most of the other plugins go to right location....11:04
sonne|workbesser82: is there  a default?11:05
sonne|workwiking: well gstreamer is stripped too11:06
besser82sonne|work: sure. cmake get's all needed defines passed during %build:11:06
besser82/usr/bin/cmake -DCMAKE_C_FLAGS_RELEASE:STRING=-DNDEBUG -DCMAKE_CXX_FLAGS_RELEASE:STRING=-DNDEBUG -DCMAKE_Fortran_FLAGS_RELEASE:STRING=-DNDEBUG -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON -DCMAKE_INSTALL_PREFIX:PATH=/usr -DINCLUDE_INSTALL_DIR:PATH=/usr/include -DLIB_INSTALL_DIR:PATH=/usr/lib64 -DSYSCONF_INSTALL_DIR:PATH=/etc -DSHARE_INSTALL_PREFIX:PATH=/usr/share -DLIB_SUFFIX=64 -DBUILD_SHARED_LIBS:BOOL=ON -DCMAKE_BUILD_TYPE=Release -DBUNDLE_EIGEN=OFF -DBUNDLE_NL11:06
besser82OPT=OFF -DBUNDLE_JSON=OFF -DENABLE_TESTING=ON -DUSE_SVMLIGHT=OFF -DCmdLineStatic=ON -DLuaModular=ON -DOctaveModular=ON -DPythonModular=ON -DRubyModular=ON ..11:06
besser82sonne|work, wiking: ^^^^11:06
sonne|workbesser82: ahh ok then it is a bug for wiking :P11:07
wikingbesser82: lib_suffix that is?11:07
besser82wiking: or "LIB_INSTALL_DIR"11:15
besser82wiking: take the one which is easier to implement for you ;)11:15
besser82wiking: shall I open an issue on gh for tracking?11:15
wikingyeah u can11:18
wikingbesser82: but LIB_INSTALL_DIR is not always set or?11:18
besser82wiking: LIB_SUFFIX either ;)11:19
wikingok then11:19
besser82wiking: i'd suggest to wrap it in some conditional... ;)11:19
besser82wiking: if NOT defined LIB_INSTALL_DIR....11:20
besser82wiking: or such...11:20
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has joined #shogun11:22
-!- mode/#shogun [+o HeikoS] by ChanServ11:22
besser82HeikoS: https://github.com/shogun-toolbox/shogun/pull/1608 && https://github.com/shogun-toolbox/shogun/issues/1604, did ya notice? ;)11:24
@HeikoSbesser82: just saw11:27
besser82HeikoS: that should do anything needed like we did yesterday ;)11:28
@HeikoScool, how to activate it?11:28
@HeikoS-DUSE_ORIGINAL_LAPACK=On ?11:28
@HeikoSand probably -DLAPACL_LIBRARIES=/for/bar ?11:28
besser82HeikoS: `cmake -DUSE_ORIGINAL_LAPACK=ON`  is enough ;)11:29
besser82HeikoS: will automatically PREPEND %{_libdir}/liblapack.so.[0-9]11:30
besser82HeikoS: ATLAS stuff will be placed after these, then11:30
besser82HeikoS: so the linker should be prefer LAPACK.so, then ;)11:31
besser82HeikoS: I might be possible we would need libclapack.so as well, but that should be handled by FindATLAS.cmake11:32
besser82HeikoS: which I can implement in the PR, too, if needed...11:32
@HeikoSokay cool11:38
@HeikoSlets wait for travis and then merge :)11:38
@HeikoSlambday: hi!11:40
lambdayHeikoS: hi11:40
lambdayHeikoS: working in notebook11:40
@HeikoSlambday: nice!11:40
@HeikoSalready got something to share? :)11:40
@HeikoSbeing curious11:40
lambdayHeikoS: took me an hour to install things properly behind proxy11:40
@HeikoSargh11:40
@HeikoSannoying11:40
lambdayHeikoS: just added up to linear solvers11:40
lambdayHeikoS: but how to port images when viewing  in nbviewer??11:41
lambdayHeikoS: yeah :( pip install doesn't work even with --proxy :(11:41
@HeikoSjust evaluate all cells, save the notebook, upload it to gist, and put the number into nbviewer11:41
@HeikoSlambday: ah stupid, well, you dont need cholmod I guess11:41
@HeikoSsince we have a linear solve using sparse LLT in shogun now11:41
@HeikoS(though its a bit slow)11:42
lambdayHeikoS: yeah we don't need cholmod11:42
lambdayHeikoS: but the images... I saved these images in a dir, from which I use markdown to show them11:43
@HeikoSlambday: what?11:44
@HeikoSlambday: which images?11:44
lambdayHeikoS: sparsity structure, eigenspectrum11:44
-!- travis-ci [~travis-ci@ec2-54-227-13-93.compute-1.amazonaws.com] has joined #shogun11:45
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1145356411:45
-!- travis-ci [~travis-ci@ec2-54-227-13-93.compute-1.amazonaws.com] has left #shogun []11:45
lambdaythe code is there but the image don't embed within it but pops out :-/11:45
@HeikoSlambday: ah11:46
@HeikoSipython notebook --pylab inline11:46
lambdayerr.. okay :D thanks :D11:47
@HeikoSlambday: or like in here11:47
@HeikoShttp://nbviewer.ipython.org/657947211:47
@HeikoS%pylab inline11:47
@HeikoSin fact, that is better11:47
lambdayHeikoS: awesome11:48
lambdaychecking out11:48
besser82wiking: there's the issue: https://github.com/shogun-toolbox/shogun/issues/160911:50
besser82wiking: as far as I can tell: libshogun.so* and lua_modular are affected11:51
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 245 seconds]12:10
-!- besser82 [~besser82@fedora/besser82] has joined #shogun12:10
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:c029:1ee2:bde3:1b86] has joined #shogun12:15
-!- mode/#shogun [+o iglesiasg] by ChanServ12:15
@iglesiasgHeikoS, hey! I am on the integration test from right now12:16
@HeikoSiglesiasg: cool! :)12:16
@iglesiasgHeikoS, I really think though that it happens because I changed the structure of LMNN (added a new member)12:17
@HeikoSiglesiasg: yep thats the reason12:17
@iglesiasgHeikoS, would that explain it?12:17
@HeikoSchanging classes breaks integration tests12:17
@HeikoS(see my issue :) )12:17
@iglesiasgHeikoS, mm ok. Does it have to be fixed manually in any case?12:17
@iglesiasgyep, I remember it12:17
@HeikoSiglesiasg: yes you have to update the file12:17
@iglesiasgHeikoS, all right, let me have a look how this works12:17
@iglesiasgI guess it should be enough removing the file12:18
@HeikoSiglesiasg: nono12:19
@HeikoSin the integration test folder in python modular12:20
@HeikoSthere are two scripts: tester.py and generator.py12:20
@HeikoStester.py <your-test-filename> runs the test12:20
@HeikoSwhich the travis does12:20
@HeikoSand generator.py creates a new file for it12:20
@HeikoSoverwriting the old12:20
@HeikoSif you did this, you have to do two commits:12:20
@HeikoSfirst commit the modified files (into the different repository shogun-data)12:21
@HeikoSsecond: commit the "data" version in the development branch12:21
@iglesiasgHeikoS, thanks for the explanation!12:23
@HeikoSiglesiasg: now you see why I find this annoying btw12:23
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has joined #shogun12:26
@iglesiasgHeikoS, question12:34
@iglesiasgHeikoS, I did the PR in shogun-data with the new file created by generator.py12:35
@HeikoSyep12:35
@iglesiasgHeikoS, and I get the message in shogun repo that the data dir has received new commits. Should I commit this version right away or wait until it is merged?12:36
@iglesiasgI have no idea really how this versions of submodule work actually12:36
@HeikoSiglesiasg: as soon as you changed/comitted them locally, you can update the data version12:37
@iglesiasgHeikoS, all right, I will make the PR right now then. If you have time, have a look12:38
@HeikoSiglesiasg: OK!12:39
@HeikoSiglesiasg: you can directly merge the data commit12:40
@HeikoSsince the main repo will only use it when you change the data version12:40
shogun-notifier-shogun: Fernando Iglesias :develop * 2e73702 / data: https://github.com/shogun-toolbox/shogun/commit/2e737021faba851a86ec6d2ce0fd719796682c4412:40
shogun-notifier-shogun: Update data version12:40
shogun-notifier-shogun: Fernando Iglesias :develop * 2b98ecb / data: https://github.com/shogun-toolbox/shogun/commit/2b98ecbf81ccc4b133ec996c2dcfd927365264eb12:40
shogun-notifier-shogun: Merge pull request #1610 from iglesias/develop12:40
shogun-notifier-shogun:12:41
shogun-notifier-shogun: Update data version12:41
shogun-buildbotbuild #2169 of deb1 - libshogun is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2169  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>12:42
shogun-buildbotbuild #2168 of deb1 - libshogun is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2168  blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com>12:42
@iglesiasgHeikoS, I think we should merge in shogun-data before merging in shogun for data version update :)12:43
shogun-notifier-shogun-data: Fernando Iglesias :master * f752c3b / testsuite/tests/metric_lmnn_modular0.txt: https://github.com/shogun-toolbox/shogun-data/commit/f752c3beae53e6eff6e22d7fa48b05086b29d04b12:43
shogun-notifier-shogun-data: New LMNN testsuite file after new member was added12:43
shogun-notifier-shogun-data: Fernando Iglesias :master * 1c7d13d / testsuite/tests/metric_lmnn_modular0.txt: https://github.com/shogun-toolbox/shogun-data/commit/1c7d13d205fba4cef831d959b625e5121cb449b712:43
shogun-notifier-shogun-data: Merge pull request #22 from iglesias/lmnn12:43
shogun-notifier-shogun-data:12:43
shogun-notifier-shogun-data: New LMNN testsuite file after new member was added12:43
@iglesiasgshogun-buildbot, force build "deb1 - libshogun"12:45
shogun-buildbotbuild forced [ETA 4m14s]12:45
shogun-buildbotI'll give a shout when the build finishes12:45
shogun-buildbotbuild #2170 of deb1 - libshogun is complete: Failure [failed configure]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/217012:45
@iglesiasgwtf12:45
@iglesiasgshogun-buildbot, force build "deb1 - libshogun"12:46
shogun-buildbotbuild forced [ETA 4m14s]12:46
shogun-buildbotI'll give a shout when the build finishes12:46
shogun-buildbotbuild #2171 of deb1 - libshogun is complete: Failure [failed configure]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/217112:46
@HeikoSiglesiasg: next thing is to push the data version12:49
@HeikoSthis fill fix the integration test then12:49
@iglesiasgHeikoS, I have pushed both to shogun and shogun-data12:50
@iglesiasgHeikoS, something is wrong in the configuration though12:50
@HeikoSah sorry did not see12:50
@HeikoSiglesiasg: what happens?12:50
@iglesiasghttp://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2171/steps/configure/logs/stdio12:50
@HeikoSah yeah buildbots do weird thing currently, we are mostly on travis12:50
@iglesiasgHeikoS, ok12:50
@HeikoSiglesiasg: I think it would be very nice to have developer howtos12:51
@HeikoSwith instructions for these kind of things12:51
@iglesiasgHeikoS, yes. The integration test part is probably explained in the integration/README though12:52
@iglesiasgthe tester.py and generator.py I mean12:52
@iglesiasgso maybe we could just add the git thing there as well12:52
@HeikoSyeah who reads that?12:52
@iglesiasghaha12:53
@HeikoSiglesiasg: how long are you around in shogun now? but never stumbled over this12:53
@HeikoSon the webpage or github, this would have been different12:53
@iglesiasgsee your point12:54
@HeikoSBut I always complain, should rather write those things12:54
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Ping timeout: 276 seconds]12:58
@iglesiasgHeikoS, so have you seen people using Julia already?12:59
@HeikoSiglesiasg: yes, one sitting in my office12:59
@iglesiasgHeikoS, cool12:59
@HeikoSiglesiasg: tried it myself, pretty nice13:01
@HeikoSI might use it in a few years :)13:01
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun13:09
@iglesiasgHeikoS, all right so lmnn integration test fixed13:11
@iglesiasga couple in svm to go13:12
@iglesiasghttps://travis-ci.org/shogun-toolbox/shogun/jobs/1145925413:12
@HeikoSiglesiasg: cool!13:12
@HeikoSiglesiasg: always be careful when overwriting the files13:12
@HeikoSits good to check whats wrong before13:12
@HeikoSif just a member was added its safe to change13:12
@iglesiasgHeikoS, aham, I understand13:12
@HeikoSbut sometimes other things are wrong (so the test actually detected an error) - though thats rare in my experience13:12
besser82HeikoS, sonne|work, wiking: Hey, guys!  Looks im heading to beta-stage with my rpms for el5, el6 and FC18+  :D13:27
@HeikoSiglesiasg: http://nbviewer.ipython.org/659306713:27
@HeikoSbesser82: awesome! :913:27
@HeikoS:)13:27
@iglesiasgHeikoS, is cholesky from numpy?13:28
besser82HeikoS, sonne|work, wiking: For el5 I'll need some stuff backported, like CMake 2.8.X, by some of my fedora-buddies. But that shouldn't be a real problem ;)13:28
@iglesiasgHeikoS, ah ok it is the PCA whitening thing13:29
@iglesiasgHeikoS, thank you!13:29
@HeikoSbesser82: nice!13:29
@HeikoSiglesiasg: yeah13:29
@HeikoSiglesiasg: sorry I imported pylab stuff13:29
@iglesiasgno problem13:29
@iglesiasgthanks13:29
@HeikoSiglesiasg: compute the covariance of the data before and after, then youll see the effect, btw here is the math: http://en.wikipedia.org/wiki/Whitening_transformation13:30
shogun-notifier-shogun: Bj?rn Esser :develop * 83e5bb9 / CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/83e5bb9f6c8c5525eb8707db031e3cfcac1d7fcc13:31
shogun-notifier-shogun: add option "USE_ORIGINAL_LAPACK" to prefer found ${LAPACK_LIBRARIES}13:31
shogun-notifier-shogun: over the ones found in ${ATLAS_LIBRARIES} during linker-stage, because13:31
shogun-notifier-shogun: some versions of liblapack.so provided by ATLAS are missing some needed13:31
shogun-notifier-shogun: symbols from the original LAPACK library -- this is case on el5, e.g.13:31
shogun-notifier-shogun: Heiko Strathmann :develop * e3e225d / CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/e3e225d6ebece678ef3cf25be4e6885f8ee761fe13:31
shogun-notifier-shogun: Merge pull request #1608 from besser82/fix_lapack_cmake13:31
shogun-notifier-shogun:13:31
shogun-notifier-shogun: add option "USE_ORIGINAL_LAPACK"13:31
besser82HeikoS: Thanks ;)13:33
shogun-buildbotbuild #2173 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2173  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>13:37
@HeikoSbesser82: trying it now13:39
besser82HeikoS: kk, go on ;)13:39
-!- lambday [67157d37@gateway/web/freenode/ip.103.21.125.55] has quit []13:46
@HeikoSbesser82: did not work13:46
@HeikoSbesser82: but getting back in an hour, going for lunch now :)13:47
besser82HeikoS: kk, can you paste the error-msg, plz?13:47
@iglesiasgHeikoS, enjoy your meal ;)13:47
shogun-buildbotbuild #2172 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2172  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>13:48
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:c029:1ee2:bde3:1b86] has quit [Read error: Connection reset by peer]13:54
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:c029:1ee2:bde3:1b86] has joined #shogun13:54
-!- mode/#shogun [+o iglesiasg] by ChanServ13:54
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has quit [Quit: Leaving.]14:02
-!- travis-ci [~travis-ci@ec2-54-227-13-93.compute-1.amazonaws.com] has joined #shogun14:09
travis-ci[travis-ci] it's Fernando Iglesias's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1145925114:09
-!- travis-ci [~travis-ci@ec2-54-227-13-93.compute-1.amazonaws.com] has left #shogun []14:09
wikingshogun-buildbot: force build --branch=develop 'deb1 - libshogun'14:19
shogun-buildbotbuild forced [ETA 4m14s]14:19
shogun-buildbotI'll give a shout when the build finishes14:19
shogun-buildbotbuild #2174 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/217414:20
@HeikoSbesser82: around?14:29
besser82HeikoS. sure!14:29
@HeikoSok then let me paste the errors14:29
@HeikoS(same as before) and the commands14:29
besser82HeikoS: same as before? Like in the issue-ticket?14:30
-!- yoo__ [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has joined #shogun14:36
yoo__hi! how to solve "g++ version is too old" ?14:36
-!- yoo__ [2eda6d58@gateway/web/freenode/ip.46.218.109.88] has quit [Client Quit]14:37
@HeikoSyoo__ use a more recent one :) you can specify another gcc/g++ bin, see readme.cmake14:40
thoralfHeikoS: yoo__ wasn't patient enough for your answer ;)14:42
-!- yo [~yoo@46.218.109.88] has joined #shogun14:44
-!- yo is now known as Guest5570314:44
Guest55703re all! how to solve "g++ version is to old ?"14:44
@HeikoSGuest55703: hi see above14:45
Guest55703don't have anythg above :)14:46
sonne|work02:40:40 PM) HeikoS: yoo__ use a more recent one :) you can specify another gcc/g++ bin, see readme.cmake14:46
sonne|work^ Guest5570314:46
-!- travis-ci [~travis-ci@ec2-54-242-16-107.compute-1.amazonaws.com] has joined #shogun14:46
travis-ci[travis-ci] it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1146060114:46
-!- travis-ci [~travis-ci@ec2-54-242-16-107.compute-1.amazonaws.com] has left #shogun []14:46
sonne|worknow you have :P14:46
Guest55703thanks!14:47
Guest55703weird, tried g++ 4.6 or clang++ 3.0 and cmake continue to raise "g++ is too old"14:51
@HeikoSGuest55703: really?14:52
@HeikoSthat should be a bug then14:52
Guest55703cmake version is 4.8.314:52
@HeikoSsince we only need 4.4 or so14:52
@HeikoScmake 4?14:52
Guest55703and each time I have remove cmake cache14:52
@HeikoS2.8.2 you mean?14:52
wikingGuest55703: what os?14:52
Guest55703ubuntu14:52
wikingGuest55703: ok try running cmake with -DENABLE_CCACHE=OFF14:53
Guest55703sss14:53
Guest55703nice it works14:53
@HeikoSwiking: I think I found a bug related to automated unit tests14:53
@HeikoSif jinja is not around there are these dummy tests14:54
wikingHeikoS: that's already fixed14:54
@HeikoSand my compiler complains about the "." in the name14:54
@HeikoS".cc_dummy"14:54
wikingHeikoS: 0d0f3eecfec68c0c09e8185f370d9092c4b4f25314:54
@HeikoSwhen was that fixed? just rebased14:54
Guest55703-DENABLE_CCACHE=OFF resolves the cmake error14:54
@HeikoSok14:54
wikingHeikoS: wtf14:54
wikingaaah i know14:54
Guest55703I dont understand why anyway14:54
wikingGuest55703: ubuntu feature14:54
wikingHeikoS: ok i'll push the fix soon14:55
@HeikoSwiking: okay thanks14:55
Guest55703ok thanks again14:55
wikingnw14:55
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has joined #shogun14:58
@HeikoSwiking: whats with this new coverage flag?15:06
wikingHeikoS: not really working yet15:06
@HeikoSwiking: ok, another question. Are debug symbols included in "Release". Or anything else that could make code (eigen3) slow?15:07
@HeikoSsince I am confused about its slow speed compared to cholmod15:08
@HeikoS(which I call from python)15:08
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has quit [Quit: Leaving.]15:16
wikingHeikoS: -g only15:31
@HeikoSwiking: google seems to be confused about whether it makes things slower or not, do you have any idea about this?15:34
@HeikoSthough  most people seem to reccomend it15:34
besser82HeikoS: in production binaries / libraries should get stripped ;)15:41
besser82HeikoS: so they don't ship any debug-symbols inside15:41
@HeikoSbesser82: what I am interested in, could that make things slow15:43
@HeikoSin particular in eigen3 with our huge huge library15:43
besser82HeikoS: yes, it can and will15:44
besser82HeikoS: there's a command called `strip` which is able to strip the library, so it will just carry ELF not DWARF15:44
besser82HeikoS: after stripping libshogun should be ~20 MByte ;)15:45
@HeikoSI see15:49
@HeikoSany idea how to do that? :)15:49
besser82HeikoS: strip -ps libshogun.so  ;)15:51
@HeikoSok cool, good to know15:51
wikingthoralf: ok so one question remains with your unit test... how to provide the proper path to data15:51
@HeikoSwiking: https://github.com/karlnapf/independent-jobs15:51
@HeikoSwiking: thats a draft for computation framework using PBS backends15:52
@HeikoSwiking: very basic, but I already used it for things15:52
@HeikoSmost interface parts are taken from lambday's implementation in shogun15:52
wikingcool15:53
wikingwill be great to include ref to it in 3.0 release15:54
thoralfwiking: I know.  Forgot to say that I had to create a symlink to make it work.15:55
wikingyeah i figured15:55
thoralfwiking: There's no unit test using data.15:55
wikingindeed15:55
wikingthoralf: you could have done the same way as i did for StreamingDenseFeatures15:55
thoralfwiking: Would it be possible that cmake defines the path and/or create a symlink?15:56
wikingthoralf: no15:56
wikingthoralf: unit tests are really not meant to read from data15:56
wikingthat's more like examples15:56
wikingso either we put this little test under examples15:56
wikingor generate the input ourselves15:56
thoralfSince the test does not depend on the content of the file, we could easily create a dummy file on-the-file with tempfile15:57
thoralfOnly needs to have > 8 lines.15:57
wikingthoralf: as i said i've done this with streamingdense15:57
thoralfwiking: Yes.15:58
wikingand i'll rename this test to streamingsparsefeatures test15:58
wikingand move it under features15:58
thoralfI don't mind.15:59
thoralfYou're test does almost the same.15:59
wikingthe question is how to generate15:59
thoralfYour15:59
wikingsparse features15:59
thoralfprint "0 1:1" 10 times? ;)16:00
wikinghehehe16:00
wikingthat's rather stupid16:00
thoralfKISS :)16:00
wikingi mean it would be better to be a bit more complex16:00
thoralfI know.  But for now it would work.16:01
thoralfCan extend is as soon as we need more.16:01
thoralfOn demand.16:01
thoralfBut I think saving sparse features is not hard either.16:01
thoralfDidn't do it so far.16:01
thoralfCSparseFeatures<sg_type>::save(CFile* writer)16:02
thoralfShould be trivial.16:02
thoralfI'll try.  Give me a second. :)16:02
wikingyeah that works16:02
wikingi'm just wondering what CFIle i should supply16:03
wikingok csvfile doesn't work16:04
wikingmaybe CLibSVMFile16:05
thoralfAsciiFile is gone?16:05
wikingthoralf: yes16:05
Guest55703ah16:06
Guest55703what if we want to save svm machine ?16:06
Guest55703SerializableAsciiFile is still available right ?16:07
wikingGuest55703: call save_serializable16:07
wikingGuest55703: yes SerializableAsciiFile is available16:07
wikingthe two things are unrelated16:07
Guest55703I ve notice SerializableAsciiFile takes sometimes long time to load, what alternative do we have ?16:08
wikingGuest55703: hdf516:08
wikingGuest55703: it's binary format16:09
wikingshould be faster16:09
thoralfwiking: LibSVMFile works.16:14
wikingthoralf: yeah i've copy pasted something16:14
wikingfrom libsvmfile unittest16:14
* thoralf created a file with 10 empty lines and it worked well (i.e. generating a segfault)16:15
thoralfwiking: We need something to create random sparse vectors.16:16
wikingthoralf: libsvmfile_unittest.cc16:17
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]16:17
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has joined #shogun16:17
-!- mode/#shogun [+o HeikoS] by ChanServ16:17
thoralfwiking: Thanks.16:18
wikingbut it's ok16:18
wikingi'm already finishing up the test16:19
thoralfwiking: Hmm.  No more code duplication.  I think it would make sense to outsource this random-matrix-stuff to somewhere.16:19
thoralfwiking: Cool.  Go ahead. :)16:19
thoralfwiking: I'm not happy with shogun messing up my /tmp: $(ls -la /tmp/ |wc -l)==15052816:20
thoralfwiking: Wouldn't it be nice to unlink when we're ready?16:21
shogun-notifier-shogun: Viktor Gal :develop * 5513e2c / tests/unit/base/clone_unittest.cc.py: https://github.com/shogun-toolbox/shogun/commit/5513e2c2ef3cadc522c91d6f53d7f251011cadbb16:24
shogun-notifier-shogun: remove .cc from dummy test function name when jinja2 is not available16:24
shogun-notifier-shogun: Viktor Gal :develop * 8abbf8b / / (3 files): https://github.com/shogun-toolbox/shogun/commit/8abbf8b5c4de579af6b90c835dc25ec698c44b6616:24
shogun-notifier-shogun: Fix StreamingSparseFeatures reading from file (issue #1583)16:24
shogun-notifier-shogun: and refactor the unit test for it so it reads from a generated file16:24
shogun-notifier-shogun: instead of data repository16:24
shogun-buildbotbuild #2176 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2176  blamelist: Viktor Gal <viktor.gal@maeth.com>16:27
wikingwtf is this16:27
shogun-buildbotbuild #2175 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2175  blamelist: Viktor Gal <viktor.gal@maeth.com>16:39
besser82HeikoS: https://github.com/shogun-toolbox/shogun/pull/161417:03
thoralfwiking: https://github.com/shogun-toolbox/shogun/pull/161517:05
-!- sonne|osx [~sonne@89.204.137.43] has joined #shogun17:17
@iglesiasgsonne|osx, hi there!17:21
@iglesiasgsonne|osx, so what is this filter thing you mentioned in here? https://github.com/shogun-toolbox/shogun-data/pull/1817:21
sonne|osxbesser82: what is https://github.com/shogun-toolbox/shogun/pull/1614/files supposed to do?17:21
sonne|osxiglesiasg: well we should write one!17:21
@iglesiasgsonne|osx, in CSVFile I guess?17:22
sonne|osxno in general17:22
sonne|osxiglesiasg: similar how python does it with gzipfile17:22
@iglesiasgsonne|osx, I see, so it could be possible to do stuff like CSVFile(InfamousFilter(gzippedFile))17:23
@iglesiasgbut it could be use somewhere else too17:23
sonne|osxsth like this yes17:23
@iglesiasgI see17:23
@iglesiasgI think that for the moment I will just extract the file using Python in my example17:24
@iglesiasgI should take care of finishing up LMNN and then the things for release so no time for this filter right now17:24
shogun-notifier-shogun: Thoralf Klein :develop * 0265621 / tests/unit/ (5 files): https://github.com/shogun-toolbox/shogun/commit/0265621379f0c9949883c55059160ee6c183dbb817:24
shogun-notifier-shogun: Added "unlink(filename)" to automatic tests for keeping /tmp clean.17:24
shogun-notifier-shogun: Thoralf Klein :develop * 74a84c0 / tests/unit/features/StreamingSparseFeatures_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/74a84c06b8c963d4af03fac5e1c84d538ed2c03c17:24
shogun-notifier-shogun: Added "unlink()" to new StreamingSparseFeatures_unittest.17:24
shogun-notifier-shogun: Soeren Sonnenburg :develop * a2a7f3f / tests/unit/ (6 files): https://github.com/shogun-toolbox/shogun/commit/a2a7f3fa159567586cb74ebc598a3a591b78a30417:24
shogun-notifier-shogun: Merge pull request #1615 from tklein23/keep_clean_tmp17:24
shogun-notifier-shogun:17:24
shogun-notifier-shogun: Added "unlink(filename)" to automatic tests for keeping /tmp clean.17:24
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.]17:25
besser82sonne|osx: On el5 atlas' lapack doesn't provide all needed symbols17:26
besser82sonne|osx: so we need to link against original LAPACK17:26
sonne|osxwiking: any clues what we can do about the build failure on deb1? It seems the breakage occurred quite a while ago. I cannot even find the last working version in waterfall ...17:27
besser82sonne|osx: but then there are some missing symbols, which are covered by CLAPACK, which is part of ATLAS17:27
sonne|osxbesser82: how do you know which one to use?17:27
besser82sonne|osx: but CLAPACK isn't looked up by FindATLAS17:27
besser82sonne|osx: so I need to use some regex magic17:28
besser82sonne|osx: because CLAPACK is ALWAYS avail from ATLAS17:28
sonne|osxbesser82: I am just trying to understand what you do...17:28
sonne|osxbesser82: so you try to link against clapack ??17:29
besser82sonne|osx: and is named like the altlas-lapack.so with an additional c inbetween17:29
besser82sonne|osx: yes17:29
besser82sonne|osx: there's need to link against clapack.so then17:29
shogun-buildbotbuild #2178 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2178  blamelist: Soeren Sonnenburg <sonne@debian.org>, Thoralf Klein <thoralf.klein@zib.de>17:30
sonne|osxbesser82: ohh your face appeared on github17:30
besser82sonne|osx: yes :)17:31
besser82sonne|osx: here's better one: https://fedoraproject.org/wiki/User:Besser8217:31
sonne|osxbesser82: anyway so in the end you just did add atlas-clapack to the linker right?17:31
sonne|osxflags I mean17:32
besser82sonne|osx: yes, that's the only thing and only if -DUSE_ORIGINAL_LAPACK=ON17:32
besser82sonne|osx: in other cases that's not needed :)17:32
besser82sonne|osx: so it won't break "regular" builds17:33
sonne|osxI wish we had some auto detection for what builds17:33
sonne|osxbut ok this is much better than before17:33
shogun-notifier-shogun: Bj?rn Esser :develop * dcd59e9 / CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/dcd59e93525e50bc00b03fe8370d1fccbd8151e517:33
shogun-notifier-shogun: add ATLAS-CLAPACK to linker, when using original LAPACK17:33
shogun-notifier-shogun: Soeren Sonnenburg :develop * b8753cb / CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/b8753cb3372d03b6661d435ef13233c879a7ce1a17:33
shogun-notifier-shogun: Merge pull request #1614 from besser82/fix_lapack_cmake17:33
shogun-notifier-shogun:17:33
shogun-notifier-shogun: add ATLAS-CLAPACK to linker, when using original LAPACK17:33
sonne|osxbesser82: I mean sty like a small code snippet that is attempted to compile / run with these flags17:33
sonne|osxs/sty/sth17:34
besser82sonne|osx: might be a possibility, but this little extra stuff would be needed on el5 or Debian Sarge or even older :)17:34
besser82sonne|osx: so it's just a small "legacy" hack17:35
-!- foulwall [~zhengyang@114.255.40.22] has quit [Ping timeout: 246 seconds]17:37
sonne|osxbesser82: well ok17:38
sonne|osxbesser82: btw, you don't happen to know a c based \compression library that operates on files and can use gzip/bzip/... underneath?17:39
besser82sonne|osx: just a second...17:39
sonne|osxbesser82: argh17:40
sonne|osxgtg17:40
sonne|osxcu17:40
-!- sonne|osx [~sonne@89.204.137.43] has quit [Quit: sonne|osx]17:40
besser82sonney2k: libarchive might be sth. you're looking for:  Libarchive is a programming library that can create and read several different streaming archive formats, including most popular tar variants, several cpio formats, and both BSD and GNU ar variants. It can also write shar archives and read ISO9660 CDROM images and ZIP archives.17:41
besser82sonney2k: It is c-compiled and can virtually read any archive/compression-format around in *nix-land :)17:42
@HeikoSbesser82: trying it out, thanks !17:44
besser82HeikoS. kk, np :)17:44
besser82HeikoS: sonne just merged it :)17:44
@HeikoSyep I saw it17:44
-!- travis-ci [~travis-ci@ec2-54-242-16-107.compute-1.amazonaws.com] has joined #shogun17:45
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1146676317:45
-!- travis-ci [~travis-ci@ec2-54-242-16-107.compute-1.amazonaws.com] has left #shogun []17:45
shogun-buildbotbuild #2179 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2179  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>17:47
@HeikoSbesser82: it seems to work fine now17:51
@HeikoScool!17:51
besser82HeikoS: :D17:51
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has joined #shogun17:54
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has joined #shogun18:02
lambdayHeikoS: hi18:02
@HeikoSlambday: hey!18:02
lambdayI have a proposal18:02
@HeikoSlambday: tell me :)18:02
shogun-buildbotbuild #2180 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2180  blamelist: Soeren Sonnenburg <sonne@debian.org>18:03
lambdayin the paper, they mentioned that for large condition number CG-M takes ages to converge... and the main motivation towards this approach is to be memory efficient for really massive matrices18:03
lambdayHeikoS: so, in the ipython demo, can I first use a really really massive matrix and show that this works, and then take ozone and show that for poor condition number convergence is really slow18:04
lambdayI mean, otherwise the main point would be missed18:04
lambdayI can use the small matrix, but I want to use a huge data for which cholesky would be just pathetic18:05
lambdayand show that for reasonable condition numbers, this indeed is a nice solution18:05
@HeikoSlambday: sorry what is the point and your plan?18:05
@HeikoSah18:05
@HeikoSyeah, the python notebooks should contain reasonable conditioned matrices18:06
lambdayalso, ozone will show the drawbacks18:06
@HeikoSyeah, but that is part of the game18:06
@HeikoSyou should definitely include it18:06
lambdaythis west0479 is really small18:06
lambdayexactly18:06
@HeikoSjust set the max iterations small and make a note that users can set larger to get more accurate results18:06
lambdaythat's what I was thinking about18:06
@HeikoSsounds good18:06
lambdayHeikoS: yeah, but for quick convergence I want to use another matrix for which our method *does* work really nicely18:07
@HeikoSyes, include a massive matrix with nice condition number18:07
lambdayHeikoS: so should I randomly generate one?18:07
@HeikoSthe west matrix is horribly conditioned, its an illustration that things even work then -- but take long18:07
@HeikoSthere should be18:07
lambdayyes18:08
@HeikoS1.) small easy matrix18:08
@HeikoS2.) huge huge easy matrix18:08
@HeikoS3.) small hard matrix (west)18:08
@HeikoS4.) large hard matrix18:08
@HeikoSand maybe one that is huge and easy *and* real life data18:08
lambdayfor large hard, I have ozone18:09
lambdaysearching for huge but easy18:09
@HeikoSlambday: we can ask daniel for that and him whether he has some nice real life PSD matrix which is huge but well conditioned18:09
@HeikoSlambday: do you want to include the ozone likelihood in your notebook?18:09
@HeikoSand pictures?18:09
@HeikoSor rather focus on the matrices?18:10
@HeikoSlambday: you can also illustrate all those solvers in the notebook18:10
lambdayHeikoS: I wish I can, but I need to understand the estimates for that18:10
lambdayHeikoS: yeah already planned18:10
lambdaythis is what I thought18:10
lambday1. introduction - theoretical background, just a bit18:10
@HeikoSlambday: I mean you can motivate the usage18:11
@HeikoSyou can use my code for the model18:11
lambday2. linear solvers18:11
@HeikoSI can also share some plotting code18:11
lambday3. logdet18:11
@HeikoSthen there is a motivation18:11
@HeikoSyou can also reference our arxiv paper and state that this is used in there18:11
wikingsonne|work: ideas wtf is happening with buildbots?18:11
lambdayHeikoS: that would be great!18:11
@HeikoSincluding the RR would be too much though18:11
@HeikoSlambday: check my github account and ozone-roulette18:11
@HeikoSthis contains code for computing the marginal likelihood for the ozone model18:11
@HeikoSand my poster contains the model itself and the log marginal likelihood18:12
lambdayalright18:12
@HeikoSyou can then say: "This can be used for performing exact MCMC on large models"18:12
@HeikoSand compute 1 estimate with low max iteration number ;)18:12
@HeikoSjust need plotting code, but will send you :)18:12
@HeikoSthen it will look very cool18:12
lambdayyeah :D18:12
@HeikoSlambday: btw try to be not tooo formal  but rather motivate things, why we want to do this etc18:13
lambdayalright..18:13
@HeikoSlambday: I think with all that and your other ideas, it will look awesome18:13
lambdaylet's see... I am hopeful :)18:14
@HeikoSlambday: https://github.com/karlnapf/ozone-roulette18:14
@HeikoSlambday: btw I also have your computation framework running in python with a batch cluster backend https://github.com/karlnapf/independent-jobs18:14
lambdayHeikoS: awesome!18:15
lambdayHeikoS: when are we going to do this inside shogun? :D18:15
@HeikoSlambday: hopefully soon18:15
@HeikoSlambday: but too many other urgent things to do currently18:15
lambdaywhen is the 3.0 release?18:16
@HeikoSlambday: and there are some hacks in there that I dont know whether its possible to do them18:16
@HeikoSlambday: two weeks or so18:16
@HeikoSlambday: we will need your help :)18:16
lambdaythe ipython notebook idea for end term is awesome by the way! really motivates the users18:16
@HeikoSlambday: yeah, that will be really useful, the three I have seen so far are very cool, waiting for yours and romans18:17
lambdayHeikoS: hehe.. :D I'll try my best18:17
@HeikoSlambday: let me know if you need any tex stuff18:17
@HeikoSlike poster sourcecode or so ....18:18
wikingmmm we have a problem with r_modular + CScalarResult<>: https://travis-ci.org/shogun-toolbox/shogun/jobs/11466773#L305518:18
wikingwhen did this change?18:18
@HeikoSthese plots that I have on the poster would be very nice18:18
lambdayHeikoS: yeah18:18
@HeikoSwiking:  sonney2k changed some R stuff18:18
lambdaywiking: last week18:18
@HeikoSah18:18
@HeikoSyou know better :)18:18
wikingHeikoS: it has nothing to do with sonney2k change18:18
lambdaywiking: checking18:19
-!- hushell [~hushell@c-98-232-178-161.hsd1.or.comcast.net] has quit [Ping timeout: 260 seconds]18:23
lambdaywiking: so far I understand, I need to change the name "result" to something else!18:31
lambdaybecause swig for R uses that same name to define something else (SEXPREC) :-/18:32
lambdayno that's not it18:34
shogun-notifier-shogun: Viktor Gal :develop * cfdbbca / / (2 files): https://github.com/shogun-toolbox/shogun/commit/cfdbbcaef92362b27cb4bf77e25185b984b9745218:35
shogun-notifier-shogun: Remove regression_svrlight_modular.cs if SVMLight is disabled (issue #1600)18:35
shogun-notifier-shogun: use appropriate macro in PermutationMatrix_unittest.cc18:35
-!- thoralf [~thoralf@enki.zib.de] has quit [Quit: Konversation terminated!]18:37
wikingbesser82: see my comments about https://github.com/shogun-toolbox/shogun/issues/160918:44
besser82wiking: use "LIB_INSTALL_DIR" then18:45
wikingbesser82: if it's set i guess..18:46
besser82wiking: with a conditional around ;)18:46
wikingbecause that's a none standard fedora hack18:46
besser82wiking: nope18:46
wikingyes18:46
besser82wiking: LIB_INSTALL_DIR18:46
wikingthere's no LIB_INSTALL_DIR in cmake itself18:46
wikingever18:46
besser82wiking: that's the usual define for where to install the libs ;)18:47
wikingusual = in case of fedora18:47
wikingnobody else uses this18:47
wikingi've been now googling around18:47
besser82wiking: SUSE, openSuse, Centos, RHEL18:47
wikingyes18:47
wikingfedora and alike18:47
wikingmeaning *rpm18:47
wikinganything that is else doesn't have it18:47
wikingarchlinux, debian, ubuntu etc18:48
besser82wiking: debain has multiarch dirs18:48
besser82wiking: like ubuntu18:48
wikingbesser82: as well as osx18:48
besser82wiking: l/usr/lib/$shitty-tuple18:48
wikingand opensolaris18:48
wikingand others18:48
lambdayHeikoS: http://www.cise.ufl.edu/research/sparse/matrices/list_by_dimension.html18:48
lambdaywiking: I am sorry I didn't get the issue :(18:48
wikinglambday: i dont know what's the problem18:49
lambdaywiking: it has to be something with some changes in swig/R in travis18:49
wikingbut the problem is that couple of days ago it compiled fine18:49
wikingno18:49
wikingi have the same on fatbot as well18:49
@HeikoSlambday: nice page! try to find a nicely conditioned one that is >30000018:49
lambdaywiking: because when I sent the PR it compiled fine18:49
wikinglambday: i highly doubt it18:49
lambdayHeikoS: checking :)18:49
wikinglambday: gimme the PR number18:49
@HeikoSlambday: sadly they did not put the condition number18:49
lambdayHeikoS: yeah :(18:50
lambdaywiking: yeah just a sec18:50
@HeikoSlambday: aim for something very sparse with a lot of structure18:50
wikinglambday: https://github.com/shogun-toolbox/shogun/pull/158618:50
wikingwas it this one?18:50
@HeikoSman these graphs look so cool!18:50
lambdayHeikoS: they sure do18:50
lambdaywiking: yes18:50
shogun-buildbotbuild #2181 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2181  blamelist: Viktor Gal <viktor.gal@maeth.com>18:50
lambday:D18:51
wikinglambday: https://travis-ci.org/shogun-toolbox/shogun/jobs/11412632#L302118:51
wikingHeikoS: ^ you've merged this breaker...18:51
wikingshogun-buildbot: force build --branch=develop 'deb1 - libshogun'18:51
shogun-buildbotbuild forced [ETA 4m14s]18:51
shogun-buildbotI'll give a shout when the build finishes18:51
wikinglambday: so it's there and it's not travis as u can see18:52
shogun-buildbotbuild #2182 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/218218:52
@HeikoSwiking: I merged since this looked like ongoing R business to me, sorry18:53
@HeikoSwiking: also isnt R excluded?18:53
lambdaywiking: I am sorry I can't get a clue18:53
@HeikoSI though is was allowed to fail18:53
lambdayHeikoS: yes R is allowed failures :D18:53
wikingit's allowed to fail18:53
wikingas we have other issues18:54
besser82wiking: where inside CMakeLists the libdir is set?18:54
@HeikoSwiking: is it really this patch that causes it?18:54
wikingHeikoS: yes18:54
@HeikoSor is it just this commit that discovered that there is a problem?18:54
wikingif u step back in travis builds one build you'll see it's examples that fails18:54
lambdayHeikoS: it has issues since in this patch I added CScalarResult and CVectorResult etc..18:54
wikingnot the interface itself18:54
lambday...to modular interface18:54
wikingbesser82: well the libshogun install 'path' is in src/shogun/CMakeLists.txt18:55
besser82wiking: thx! will have a look on it ;)18:55
lambdaywiking: HeikoS: R uses some internal structure SEXPREC18:56
besser82wiking: the ruby thing I already found and can change it with a sed-one-liner in spec-file ;)18:56
@HeikoSwiking: but thats an R issue isnt it?18:56
lambdaywhich is incomplete, don't know why18:56
@HeikoSI have no idea about this stuff18:56
lambdayneither do I18:56
@HeikoSwhy is it caused by these classes18:56
wikingbesser82: i'll add a test for LIB_INSTALL_DIR18:56
@HeikoSthey are very simple18:56
wikingbesser82: if it's there then libshogun is going to be installed in LIB_INSTALL_DIR18:56
wikingok?18:56
besser82wiking: that would be really nice of you :)18:57
besser82wiking: for lua, too, then?18:58
shogun-notifier-shogun: Viktor Gal :develop * 59b54f6 / src/shogun/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/59b54f65a6855406db26ffe2ac305ec72facc20e18:59
shogun-notifier-shogun: Towards fixing issue #160918:59
shogun-notifier-shogun: install libshogun to LIB_INSTALL_DIR if it's set18:59
wikingbesser82: have u checked how lua install path is defined?18:59
besser82wiking: yes :)19:00
besser82wiking: LIB_SUFFIX might be useful there ;)19:00
wikingbesser82: i'll just use LIB_INSTALL_DIR19:00
wikingand then just put there that into the regexp if it's set19:01
besser82wiking: or so ;)19:01
besser82wiking: I'm fine with either :)19:01
wikingshogun-buildbot: force build --branch=develop 'deb1 - libshogun'19:03
shogun-buildbotbuild forced [ETA 4m14s]19:03
shogun-buildbotI'll give a shout when the build finishes19:03
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 245 seconds]19:06
shogun-notifier-shogun: Viktor Gal :develop * 473ed06 / src/interfaces/lua_modular/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/473ed06e29d56f240a4f296da4d4f6ab36a4fa9319:08
shogun-notifier-shogun: Install lua modular interface to under LIB_INSTALL_DIR if it's set19:08
shogun-notifier-shogun: issue #160919:08
shogun-buildbotHey! build deb1 - libshogun #2184 is complete: Exception [exception]19:10
shogun-buildbotBuild details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/218419:10
besser82wiking: thanks a lot ;)19:10
wikingsonne|work: i think /tmp is full on buildslave419:14
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has joined #shogun19:16
-!- mode/#shogun [+o HeikoS] by ChanServ19:16
shogun-buildbotbuild #2185 of deb1 - libshogun is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2185  blamelist: Viktor Gal <viktor.gal@maeth.com>19:16
shogun-buildbotbuild #2183 of deb1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2183  blamelist: Viktor Gal <viktor.gal@maeth.com>19:18
@HeikoSbesser82: around?19:19
@HeikoSI have run into an annoying problem19:19
@HeikoSwhich is: shogun needs classic lapack, and python needs atlas lapack19:19
-!- lisitsyn [~lisitsyn@fb2-lo1.global63.net] has joined #shogun19:20
@HeikoSdepending which path is first in LD_LIBRARY_PATH, one of them works the other doesnt19:20
@HeikoSbut since I want to use shogun AND scipy from the same python session, I dont know what to do19:21
wikingshogun-buildbot: force build --branch=develop 'deb1 - libshogun'19:22
shogun-buildbotbuild forced [ETA 4m14s]19:22
shogun-buildbotI'll give a shout when the build finishes19:22
@HeikoSwiking: any ideas?19:23
@HeikoS^19:23
shogun-buildbotbuild #2186 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/218619:25
wikingyeeey buildbot back in business19:25
besser82HeikoS: sure! go ahead :)19:27
@HeikoSbesser82:  see above19:27
besser82HeikoS: btw. cuold you merge, plz?  ---> https://github.com/shogun-toolbox/shogun/pull/161619:27
besser82HeikoS: will investigate your issue meanwhile ;)19:28
besser82HeikoS: do have any error messages for me?19:28
shogun-notifier-shogun: Bj?rn Esser :develop * 0a92dab / examples/undocumented/matlab_and_octave/tests/ (2 files): https://github.com/shogun-toolbox/shogun/commit/0a92dab1d98340073e53aa2c56b28c29bbe5366319:28
shogun-notifier-shogun: get rid of rpmlint "W: wrong-file-end-of-line-encoding"19:28
shogun-notifier-shogun: Bj?rn Esser :develop * ab0edc7 / src/shogun/ (10 files): https://github.com/shogun-toolbox/shogun/commit/ab0edc793de5b9143c2b3d2b96ef6509ce86307619:28
shogun-notifier-shogun: get rid of rpmlint "E: incorrect-fsf-address"19:28
shogun-notifier-shogun: Bj?rn Esser :develop * 198eb1c / examples/descriptions/ (3 files): https://github.com/shogun-toolbox/shogun/commit/198eb1c3c131a0fc149865140827d92c696ce60a19:28
shogun-notifier-shogun: get rid of rpmlint "W: file-not-utf8"19:28
shogun-notifier-shogun: Heiko Strathmann :develop * 9bae02f / / (15 files): https://github.com/shogun-toolbox/shogun/commit/9bae02f15beed080e023bd2137912c5d1557ccd519:28
shogun-notifier-shogun: Merge pull request #1616 from besser82/fix_encoding19:28
shogun-notifier-shogun:19:28
shogun-notifier-shogun: get rid of some rpmlint issues19:28
@HeikoSbesser82: well, symbol not found19:28
@HeikoSeither from running shogun unit-tests19:29
@HeikoSor python -c 'from scipy.linalg.basic import solve_triangular'19:29
besser82HeikoS: sth. more detailed?19:29
@HeikoSok19:29
besser82HeikoS: like stderr?19:29
@HeikoS[ RUN      ] LogDetEstimator.sample_ratapp_probing_sampler19:29
@HeikoS./shogun-unit-test: symbol lookup error: /home/ucabhst/shogun/build/src/shogun/libshogun.so.14: undefined symbol: dstemr_19:29
@HeikoSthats the one from shogun-unit-tests19:30
@HeikoSImportError: /share/apps/python-2.7.1/lib/python2.7/site-packages/scipy/linalg/clapack.so: undefined symbol: clapack_sgesv19:30
@HeikoSand thats the one from the python import19:30
@HeikoSso it turns out that scipy needs the atlas-lapack version19:30
besser82ldd $shogun.so ?19:30
@HeikoSwhere my shogun can only work with the classic one19:30
besser82HeikoS: ldd $shogun.so ?19:30
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has quit [Excess Flood]19:31
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has joined #shogun19:32
-!- mode/#shogun [+o HeikoS] by ChanServ19:32
@HeikoSbesser82: did you get that?19:32
@HeikoSwas kicked due to flooding19:32
besser82HeikoS: not fully :(19:32
@HeikoSok let me pastebin19:32
besser82HeikoS: would be an idea :)19:32
@HeikoShttp://pastebin.com/LEYz0FiL19:33
shogun-buildbotbuild #2187 of deb1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/2187  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>19:33
@HeikoSbesser82: and with19:34
@HeikoS/usr/lib64/atlas:/usr/lib64:/home/ucabhst/shogun-install/lib:/opt/gridengine/lib/lx26-amd64, python works but shogun doesnt19:35
@HeikoSand /usr/lib64:/home/ucabhst/shogun-install/lib:/opt/gridengine/lib/lx26-amd64 does vice versa with the errors I sent19:35
-!- zxtx [~zv@ip-64-134-197-230.public.wayport.net] has quit [Ping timeout: 248 seconds]19:36
@HeikoSbesser82: and obviously the ldd changes if the LD_PATH changes19:37
besser82HeikoS. I know ;)19:37
@HeikoSbesser82: I could get around this via exposing a triangular solver to the modular interfaces, but this seems not good19:37
@HeikoSsince I cannot use any scipy stuff with shogun then19:37
besser82HeikoS: Let me investigate and solve ;)19:38
@HeikoSbesser82: can one merge two shared libs?19:38
@HeikoSthen I can put that merged thing into my home and link against it19:38
besser82HeikoS: that's not possible, unfortunately19:39
@HeikoSscipy seems to be picky and only accepts the atlas-lapack19:39
@HeikoSshogun is picky and only takes my non-atlas lapack19:39
besser82HeikoS: possibly, but nothing which can't be solved ;)19:39
@HeikoShehe I like your approach19:40
@HeikoSgetting some coffee ...19:40
besser82HeikoS: kk :)  Thanks for the compliment!19:40
shogun-buildbotbuild #1743 of bsd1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/174319:41
besser82HeikoS: can you please paste the output of `yum deplist scipy` ?19:42
besser82HeikoS: on fpaste || pastebin?19:42
@HeikoSthats empty19:43
@HeikoSscipy is installed to /shape/apps19:43
besser82HeikoS: yes, el5, forgot...19:43
shogun-buildbotbuild #2188 of deb1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb1%20-%20libshogun/builds/218819:43
besser82HeikoS: the the problem is that ver of scipy is not properly configured :(19:44
besser82HeikoS: at least it seems...19:44
@HeikoSbesser82: ok, what should I tell the guy?19:44
@HeikoSbesser82: or can I install locally instead?19:44
@HeikoScompile locally or so?19:44
@HeikoSor even virtual-env?19:44
besser82HeikoS: might be one option, or let me build some backported rpm ;)19:44
@HeikoSthe guy told me I dont need to since he can do things, but since nothing works that might be an option19:45
shogun-buildbotbuild #1744 of bsd1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/1744  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>19:45
@HeikoSbesser82: can I install that even?19:45
besser82HeikoS: no, but your admin should :)19:45
@HeikoSbesser82: I dont think he likes that19:45
@HeikoSbesser82: what about just virtual-enving a full python distribution?19:45
besser82HeikoS: would be the other possibility19:46
besser82HeikoS: but your admin can fetch my build from fedora-infrastructure ;)19:46
@HeikoSbesser82: or just scipy19:46
besser82HeikoS: possibly scipy might be enough19:46
@HeikoSmmmh dont know I dont like to interact since things take so long19:46
besser82HeikoS: but let me check first for a simpler approach :)19:46
@HeikoSso try-error is annoying this way19:47
besser82HeikoS: yes19:47
besser82HeikoS: why you admin's don't use lxc? :D19:47
besser82HeikoS: would make thind real easy for all ;)19:47
besser82s/thind/things19:47
@HeikoSbesser82: dont ask, the worst thing is that this is different for every computer you touch19:47
besser82HeikoS: phuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu!19:48
@HeikoSvirutal-env is not installed19:49
@HeikoSman19:49
@HeikoSI dont know how long I should spend with this any longer19:49
@HeikoSjust installing software19:49
@HeikoSbesser82: whats your suggestion?19:50
besser82chroot in your home :)19:50
@HeikoS?19:50
wikingHeikoS: yeah why dont u prepare a custom slim os and chroot it19:51
shogun-buildbotbuild #97 of FC19 - modular_interfaces is complete: Failure [failed configure]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20modular_interfaces/builds/97  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Bj?rn Esser <bjoern.esser@gmail.com>19:51
@HeikoSbesser82, wiking because I dont know about these things, how would that work?19:51
wikingHeikoS: ok dont worry19:52
wikingHeikoS: hopefully tonight i'll get ourselves a virtualenved image19:52
wikingand then you can use that anywhere19:52
wiking:)19:52
wikingHeikoS: that's for ipython notebook19:52
besser82HeikoS: I can give you a pristine fedora-chroot19:52
@HeikoSwiking: wow, were you doing that anyways=?19:52
@HeikoSah19:52
besser82HeikoS: just un-tar && go ;)19:52
@HeikoSso then I can just somehow open this and have everything there?19:52
wikingHeikoS: yes19:52
@HeikoSwhat about paths/libs dependencies etc?19:53
wikingHeikoS: it's within that19:53
@HeikoSis it fully self-contained?19:53
wikingHeikoS: yes19:53
@HeikoSwhat does it need then?19:53
wikingHeikoS: just pray that u dont have harsh quota-s on your home19:53
@HeikoSwiking: what about thinkgs like scipy?19:53
@HeikoSthats not needed for shogun19:53
@HeikoSand other software19:53
@HeikoScan one install things?19:53
wikingHeikoS: it's needed for python_modular examples19:53
wikingso it'll be there19:54
@HeikoSand scikits.cholmod?19:54
wikingHeikoS: i'm planning to add this http://thomassileo.com/blog/2012/11/19/setup-a-remote-ipython-notebook-server-with-numpyscipymaltplotlibpandas-in-a-virtualenv-on-ubuntu-server/19:54
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has quit [Quit: Page closed]19:54
@HeikoSwiking: ok this stuff would actually be enough for me19:54
wikingHeikoS: what other sw woudl u need?19:54
@HeikoSwiking: let me check19:55
@HeikoSwiking: how about pulling/rebuilding shogun?19:55
wikingif u need any other image let me know19:55
wikingHeikoS: ?19:55
wikingHeikoS: that's another story19:55
wikingi'll prepare the image + a shogun in it19:55
@HeikoSwiking: like I make changes to the development branch19:55
wikingyeah i know u want everything19:55
wikingthere'll be a shogun git repo in that image19:56
@HeikoSwiking: just asking ;)19:56
wikingand u do with that whatever u want19:56
wikingbut in our case it'll be a nightly updated image19:56
@HeikoSwhat about git pull and makeing?19:56
wikingsince i want to mount that image with R-O mode19:56
wikingi mean R/O mode19:56
@HeikoSah19:56
@HeikoSI see19:56
@HeikoSfor security stuff19:56
wikingyes19:56
@HeikoSwiking: ok that is not enough for me, sadly, maybe later once things are stable19:57
wikingbut then again u do whatever u want19:57
wikingi mean u can do whatever with that image19:57
wikingand modify it as u wish19:57
@HeikoSwiking: so I could mount rw and then git pull blabla19:57
wikingand then just distrib that image around19:57
wikingyes19:57
wikingwhat u do with that image19:57
wikingis really up to you19:57
wikingthe repo with a dev branch in it will be available19:58
wikingand it'll have all the required packages19:58
wikingapart from like mosek19:58
@HeikoSwiking: so also compiler etc?19:58
wikingyes19:58
wikingas it needs it for building19:58
wikingi mean i'll rebuild the shogun in that image every night19:58
wikingbut you rebuild it whenever u want it19:59
@HeikoSwiking: hey man, that sounds too good to be true19:59
wikingwe have such image based on fedora 1919:59
wikingout of box19:59
@HeikoSI already spent a week on getting all this shit to work19:59
wikingso if u dont have a problem with fedora19:59
wikingi can tar gz u that image :P19:59
@HeikoSwiking: I dont care at all about internals, just care about things working ;)19:59
wikingHeikoS: it's 12 gigs20:00
@HeikoSwiking: thats fine20:00
wikingHeikoS: but i'll try to make a slim one assap20:00
wikingwe can even distrib this image20:00
wikingas like20:00
wikinghere u go20:00
wikinghave this as a VM image20:01
@HeikoSwiking: so how certain are you about tonight about this?20:01
wikingand you have shogun out of box20:01
@HeikoSwiking: yeah very nice20:01
wikingwell i'll have a coffe now20:01
wikingand i'll start20:01
@HeikoSwiking: ah there is a trick20:01
@HeikoSthe cluster software20:01
wikingwhat's with it?20:01
@HeikoSthe SGE stuff20:01
@HeikoScan I still run commands like qsub?20:01
wikingwell20:01
wikingis there a package for it?20:01
wikingor that's some proprietary stuff?20:02
@HeikoSwiking: no since its all installed locally and I have no idea20:02
wikingok what is the cluster stuff u r using?20:02
wikingbut i mean man how the fuck is it possible that u have a cluster20:02
wikingand the nodes are not the exact same distribs20:02
@HeikoSSun stuff20:02
@HeikoSwiking: thats solved20:02
wikingwho to the hell maintains such cluster?20:02
wikingwhere nodes are like 'whatever'20:03
@HeikoSI just want to install on my login node currently20:03
wikingok20:03
wikingi'll start up a new image20:03
@HeikoSwiking: dude I dont know how can that be, I spend days over days on this shit, totally stresses me out20:03
@HeikoSall these weird problems20:03
wikingyeah20:03
@HeikoSwhy the hell is that stupid symbol not in atlas-lapack20:03
@HeikoSI dont know20:03
@HeikoSwhy doesnt scipy work with the classic lapack20:04
@HeikoSI dont know20:04
wikinghehe20:04
wikingok20:04
wikingi'll give u a ping as soon as i'm ready20:04
@HeikoSwiking: cool cheers20:04
wikingi'll try to find an ubuntu chroot image20:04
wikingand then setup everything in it20:04
besser82HeikoS: that stupid symbol isn't in THAT particular version of atlas :)20:04
besser82HeikoS: on el6 it is :)20:05
wikingfuck it20:06
wikingit's going to be archlinux20:06
wiking:)20:06
@iglesiasgyaay! Tapkee is officially published now!!20:06
wikingiglesiasg: ^ agree? :)20:06
@iglesiasghttp://jmlr.org/papers/v14/lisitsyn13a.html20:06
@iglesiasgwiking, sure! :)20:06
wikingcongratz20:06
wikingi've put archlinux on my raspberry pi20:06
wikingand it's really cool20:06
@iglesiasglisitsyn, vodking time20:06
@iglesiasgwiking, nice! what are you doing with it?20:07
wikingiglesiasg: usual shit at the moment20:07
wikingmedia server20:07
wiking+ measuring pool temperature20:07
@iglesiasghehe20:07
-!- HeikoS1 [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has joined #shogun20:07
wiking+ gps20:07
@iglesiasgwiking, do you have a pool man?20:07
wikingand now i'm trying to do water clorization20:07
wikingiglesiasg: not mine20:07
wikingiglesiasg: it's my wife's parents house20:07
wikingthey do have a pool20:08
wiking:)20:08
shogun-buildbotbuild #1196 of rpm1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/rpm1%20-%20libshogun/builds/1196  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>20:08
@iglesiasgaah I see. Awesome anyway -- you get to bath there :D20:08
wikinganyhow20:08
wikingi'm going to get now a coffee20:08
lisitsynha20:08
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 260 seconds]20:09
wikingand then i'll see wtf is with this ipython image20:09
-!- lambday [67157f37@gateway/web/freenode/ip.103.21.127.55] has joined #shogun20:11
HeikoS1wiking: I just wrote the guy to ask whether one can submit jobs from another computer, lets see what he says20:11
HeikoS1wiking: installing python locally in the meantime20:12
lambdayHeikoS1: found  a 715,176x715,176 spd, nnz=4,817,870, nnz(cholesly)=177,279,446 (with AMD), max/min eigvals=91873.0503669/0.0299489432691, colors required=1320:14
lambdayLanczos converges within ~2000 iterations20:14
HeikoS1lambday: sweet! :) how long does that take? How long does the family solver take?20:14
lambdayHeikoS1: checking20:15
lambdayCG-M did not converge within default (1000) iterations though20:16
HeikoS1at least 2000 I guess20:16
lambday~10^6 condition number here too :(20:16
HeikoS1lambday: better than 10^1020:17
HeikoS1lambday: could also add a ridge on the orig matrix20:17
lambdayHeikoS1: that would help a bit ain't it20:18
HeikoS1lambday: try it20:18
shogun-buildbotbuild #1478 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1478  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Bj?rn Esser <bjoern.esser@gmail.com>20:18
lambdaychecking20:19
lambdayerr stupid me! was using A'A+I code for this one too argh!20:19
HeikoS1?20:19
lambdaywas trying using the pymodular script that I added20:20
lambdayjust changed the name but forgot to use actual matrix, instead was computing A^TA+I that was required for west047920:20
HeikoS1lambday: I see20:21
HeikoS1so what is the new one doing?20:21
HeikoS1lambday: actually doesnt matter if it stays sparse and is nice :)20:21
lambdaygotta check.. new one is already spd20:21
-!- hushell [~hushell@8-92.ptpg.oregonstate.edu] has joined #shogun20:23
shogun-buildbotbuild #1825 of deb3 - modular_interfaces is complete: Failure [failed compile r_modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1825  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Bj?rn Esser <bjoern.esser@gmail.com>20:26
wikingHeikoS1: imho this wont be good for u20:28
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has quit [Quit: Leaving.]20:29
HeikoS1wiking: at least now shogun AND scipy work, just tried it20:29
@iglesiasgall right, dinner time. See you later!20:29
wikingbecause the problem is that u need some stuff to mount20:29
wikingand for that u would need root privilages20:29
HeikoS1ah20:29
HeikoS1......20:29
HeikoS1well20:29
HeikoS1then virtual-env20:30
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:c029:1ee2:bde3:1b86] has quit [Quit: Ex-Chat]20:30
shogun-buildbotbuild #1826 of deb3 - modular_interfaces is complete: Failure [failed compile r_modular]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1826  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>20:36
wikinglambday: ^ same error on the buildbot20:38
lambdaywiking: HeikoS1: I guess I should remove the computation framework related things from modular interface for the time being :(20:40
HeikoS1lambday: nope20:40
HeikoS1I need it ;)20:40
lambdayHeikoS1: would you need the Scalar/Vector-results related stuffs too?20:41
HeikoS1lambday: dont know, at least the serial computation engine, I dont know if things still work when the others are removed, but you can try20:41
HeikoS1but pls do that locally first20:41
lambdayHeikoS1: alright20:44
lambdayHeikoS1: LOL I added  a ridge... 10,000 (just for fun), and CG-M converges within <50 iterations :D20:45
HeikoS1lambday: hehe :)20:45
lambdaycond# for this one is ~10020:45
HeikoS1lambday: this makes it almost a diagonal matrix20:45
HeikoS1so maybe play with that a big20:45
lambdayyeah20:45
HeikoS1lambday: what happened to the original matrix without AtA?20:45
lambdaytrying with various combo20:46
lambdayHeikoS1: going to get tea (chai) :)20:46
HeikoS1enjoy :)20:46
HeikoS1I will leave here soon too20:46
lambdayalright - I will try to do as much as possible but this stupid LAN ban thingi.. can't get any work done from 1:30AM to 5AM :'(20:47
lisitsynhah yeah lambday that's ridiculous20:51
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has joined #shogun20:54
-!- mode/#shogun [+o HeikoS] by ChanServ20:54
-!- HeikoS1 [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has quit [Read error: Connection reset by peer]20:54
@HeikoSlambday: ah man, weird ;) so gogogogog21:00
shogun-buildbotbuild #1479 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1479  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>21:01
lambdayback21:11
lambdaylisitsyn: HeikoS: tell me about it..21:11
lambdayweirdest thing I have ever heard/been through!21:12
@HeikoSlambday: I just got the log-det sampler running onthe cluster for the first time :)21:14
lambdayHeikoS: yippie :D21:15
besser82HeikoS: small fixup: https://github.com/shogun-toolbox/shogun/pull/1617 --> Can you merge, please?21:15
lambdayHeikoS: which data you are using21:15
@HeikoSlambday: ozone21:16
lambdaynice!!21:16
shogun-notifier-shogun: Bj?rn Esser :develop * 2c3c872 / src/shogun/mathematics/munkres.cpp,src/shogun/mathematics/munkres.h: https://github.com/shogun-toolbox/shogun/commit/2c3c872c5e264c547dec227d97ea93252a45237f21:16
shogun-notifier-shogun: fixed wrong license-disclaimer21:16
shogun-notifier-shogun: Heiko Strathmann :develop * 33c140f / src/shogun/mathematics/munkres.cpp,src/shogun/mathematics/munkres.h: https://github.com/shogun-toolbox/shogun/commit/33c140f53473841a79b0dea53efa87dfd27d404821:16
shogun-notifier-shogun: Merge pull request #1617 from besser82/fix_license21:16
shogun-notifier-shogun:21:16
shogun-notifier-shogun: fixed wrong license-disclaimer21:16
@HeikoSthanks besser8221:16
besser82HeikoS: np :)21:18
lambdayHeikoS: added a ridge 100.0 to the original matrix, CG-M converges within 500 iterations21:19
lambdaytested with 1000.0, 10000.0 too, CG-M converged < 50 iterations21:19
@HeikoSlambday: ok that is good, you can add a comment in the notebook that you are doing this in order to make runtime reasonable21:20
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has left #shogun []21:20
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has joined #shogun21:20
-!- mode/#shogun [+o HeikoS] by ChanServ21:20
lambdayHeikoS: yeah that's what I was thinking21:20
@HeikoSlambday: ok, could you send me a little draft of the notebook tonight?21:21
@HeikoSlambday: I have to go now, but very curious to have a look21:21
shogun-buildbotbuild #1827 of deb3 - modular_interfaces is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1827  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>21:24
shogun-buildbotbuild #1713 of deb2 - static_interfaces is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb2%20-%20static_interfaces/builds/1713  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>21:25
shogun-buildbotbuild #1828 of deb3 - modular_interfaces is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1828  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>21:25
lambdayHeikoS: currently adding/removing a lot of stuffs... will send by tomorrow?21:27
@HeikoSlambday: it doesnt have to be perfect, I am just interested how it looks like, so non-perfect state is ok21:27
lambdayHeikoS: alright I'll send :)21:28
shogun-buildbotbuild #1745 of bsd1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/174521:29
@HeikoSlambday: nice, ok see you later going home now21:31
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.]21:31
shogun-buildbotbuild #98 of FC19 - modular_interfaces is complete: Failure [failed git]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20modular_interfaces/builds/98  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Bj?rn Esser <bjoern.esser@gmail.com>21:32
besser82github is down :(21:35
-!- HeikoS [~heiko@nat-179-139.internal.eduroam.ucl.ac.uk] has joined #shogun21:37
-!- mode/#shogun [+o HeikoS] by ChanServ21:37
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun21:39
-!- mode/#shogun [+o iglesiasg] by ChanServ21:39
@iglesiasghi hushell21:40
@iglesiasghushell, how are you? I am checking the PR right now21:40
-!- travis-ci [~travis-ci@ec2-50-17-177-60.compute-1.amazonaws.com] has joined #shogun21:43
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1147004721:43
-!- travis-ci [~travis-ci@ec2-50-17-177-60.compute-1.amazonaws.com] has left #shogun []21:43
wikingshouldn't we start supporting this: http://www.openblas.net/?21:50
@HeikoSwiking: no we should stick with eigen3, which in fact calls blas internally quite often21:55
@HeikoSwiking: btw about the -g, eigen3 sometimes seems unreasonably slow, it might have to do with that, is it possible to add a strip option to cmake=?21:56
-!- lambday [67157f37@gateway/web/freenode/ip.103.21.127.55] has quit [Ping timeout: 250 seconds]21:56
wikingHeikoS: no you dont need cmake to strip the library21:57
wikingjust exec strip libshogun.so...21:57
@HeikoSwiking: can I just do this on the installed one?21:57
@HeikoSI see21:57
besser82HeikoS: yes, you can simply strip the installed one ;)21:57
besser82wiking: but it would be nice if CMake would place the system's CFLAGS/LDFLAGS at the end of all used flags ;)21:58
shogun-buildbotbuild #1480 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1480  blamelist: Bj?rn Esser <bjoern.esser@gmail.com>21:59
besser82HeikoS: strip -pg $installed_lib21:59
-!- travis-ci [~travis-ci@ec2-50-17-177-60.compute-1.amazonaws.com] has joined #shogun22:01
travis-ci[travis-ci] it's Soeren Sonnenburg's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1146952222:01
-!- travis-ci [~travis-ci@ec2-50-17-177-60.compute-1.amazonaws.com] has left #shogun []22:01
-!- travis-ci [~travis-ci@ec2-107-21-194-145.compute-1.amazonaws.com] has joined #shogun22:15
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1147269722:15
-!- travis-ci [~travis-ci@ec2-107-21-194-145.compute-1.amazonaws.com] has left #shogun []22:15
-!- FSCV [~FSCV@216-230-229-167-colo.oplink.net] has joined #shogun22:24
shogun-buildbotbuild #1197 of rpm1 - libshogun is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/rpm1%20-%20libshogun/builds/119722:29
shogun-buildbotbuild #1481 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1481  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>22:36
wikinglisitsyn: here?22:37
lisitsynwiking: ja22:37
wikinglisitsyn: what do you think about supplying a DO custom droplet for shogun? :)22:37
lisitsynwiking: haha why?22:37
wikinglisitsyn: well maybe we could run there that ipython shit22:38
lisitsynwiking: ahh well22:38
lisitsynI don't mind22:38
lisitsyn:)22:38
wikinganyways soon i have the squashfs ready22:39
wikingit's like minutes22:39
shogun-notifier-shogun: Viktor Gal :develop * 1e7792f / cmake/FindARPACK.cmake: https://github.com/shogun-toolbox/shogun/commit/1e7792f38a19d1ba655a73cfabc6574588abf61522:48
shogun-notifier-shogun: Prefer arpack over parpack22:48
-!- travis-ci [~travis-ci@ec2-50-19-39-224.compute-1.amazonaws.com] has joined #shogun22:59
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1147370622:59
-!- travis-ci [~travis-ci@ec2-50-19-39-224.compute-1.amazonaws.com] has left #shogun []22:59
shogun-buildbotbuild #99 of FC19 - modular_interfaces is complete: Failure [failed configure]  Build details are at http://buildbot.shogun-toolbox.org/builders/FC19%20-%20modular_interfaces/builds/99  blamelist: Viktor Gal <viktor.gal@maeth.com>23:01
shogun-buildbotbuild #1714 of deb2 - static_interfaces is complete: Success [build successful]  Build details are at http://buildbot.shogun-toolbox.org/builders/deb2%20-%20static_interfaces/builds/171423:02
shogun-buildbotbuild #1198 of rpm1 - libshogun is complete: Failure [failed compile]  Build details are at http://buildbot.shogun-toolbox.org/builders/rpm1%20-%20libshogun/builds/1198  blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>23:05
shogun-buildbotbuild #1482 of cyg1 - libshogun is complete: Failure [failed test]  Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1482  blamelist: Viktor Gal <viktor.gal@maeth.com>23:31
@sonney2kbesser82, libarchive sounds great!23:36
besser82sonney2k: then I'm glad to have you helped-out :)23:37
lisitsynwiking: around?23:37
besser82sonney2k: any idea why ctest doesn't respect PYTHONPATH?23:37
wikinglisitsyn: yes23:38
lisitsynwiking: have you ever worked with detection of multiple parts?23:38
wikingy23:38
besser82sonney2k,wiking: any idea why ctest doesn't respect PYTHONPATH?23:38
lisitsynwiking: I know latent svm is the approach but I am out of time :(23:38
wikinghttp://maeth.com/shogun-image.fs.sfs23:38
wikingshogun prebuild squashfs23:38
besser82wiking: any idea why ctest doesn't respect PYTHONPATH?23:39
lisitsynwiking: what have you used to learn the model, etc?23:39
wikinglisitsyn: how many parts?23:40
wikinglisitsyn: do u have the parts annotated?23:40
lisitsynwiking: 1+1 optional23:40
wikingbesser82: dunno23:40
besser82wiking: kk23:40
lisitsynwiking: basically whether a payloader carries something23:40
wikinglisitsyn: do u have annotation?23:40
lisitsynwiking: yeah I annotated a bit23:40
wikingok23:40
wiking1+1 optional is really not much23:41
wikingi mean 1 part is like no part23:41
lisitsynyeah totally simple23:41
lisitsynyes23:41
wikingtry the partbaseddetector23:41
wikingdo u have a matlab around?23:41
lisitsynnope23:41
wikinglisitsyn: https://github.com/wg-perception/PartsBasedDetector23:42
wikingbut the training part is matlab23:42
lisitsynwiking: ok thanks23:42
wikingwhere should we put a link for this squashfs23:43
wiking?23:45
lisitsynno clue :)23:45
-!- travis-ci [~travis-ci@ec2-107-21-194-145.compute-1.amazonaws.com] has joined #shogun23:51
travis-ci[travis-ci] it's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/1147406323:51
-!- travis-ci [~travis-ci@ec2-107-21-194-145.compute-1.amazonaws.com] has left #shogun []23:51
--- Log closed Wed Sep 18 00:00:04 2013