--- Log opened Fri Sep 20 00:00:06 2013 | ||
shogun-buildbot | build #1501 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1501 blamelist: Soeren Sonnenburg <sonne@debian.org> | 00:06 |
---|---|---|
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has quit [Quit: Leaving.] | 00:07 | |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 14914b9 / examples/undocumented/python_modular/kernel_rationalquadratic_modular.py: https://github.com/shogun-toolbox/shogun/commit/14914b969758afa9dc40e16dc918a4971b1bc780 | 00:21 |
shogun-notifier- | shogun: fix rational quadratic | 00:21 |
-!- travis-ci [~travis-ci@ec2-23-22-84-25.compute-1.amazonaws.com] has joined #shogun | 00:40 | |
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/11569862 | 00:40 |
-!- travis-ci [~travis-ci@ec2-23-22-84-25.compute-1.amazonaws.com] has left #shogun [] | 00:40 | |
shogun-buildbot | build #1502 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1502 blamelist: Soeren Sonnenburg <sonne@debian.org> | 01:06 |
shogun-buildbot | build #1850 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1850 | 01:35 |
-!- travis-ci [~travis-ci@ec2-23-22-84-25.compute-1.amazonaws.com] has joined #shogun | 01:44 | |
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/11571778 | 01:44 |
-!- travis-ci [~travis-ci@ec2-23-22-84-25.compute-1.amazonaws.com] has left #shogun [] | 01:44 | |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 256 seconds] | 02:05 | |
shogun-buildbot | build #19 of clang34 - static analysis is complete: Failure [failed analyse] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/19 blamelist: Soeren Sonnenburg <sonne@debian.org> | 02:20 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun | 02:26 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 02:26 | |
shogun-buildbot | build #7 of clang34 - undefined behaviour analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/7 blamelist: Soeren Sonnenburg <sonne@debian.org> | 02:30 |
-!- pickle27 [~Kevin@208.138.34.236] has joined #shogun | 02:36 | |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Read error: Operation timed out] | 02:38 | |
-!- pickle27 [~Kevin@208.138.34.236] has quit [Quit: Leaving] | 02:42 | |
shogun-buildbot | build #11 of clang34 - thread analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/11 blamelist: Soeren Sonnenburg <sonne@debian.org> | 02:45 |
shogun-buildbot | build #473 of nightly_none is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_none/builds/473 | 03:09 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 03:21 | |
shogun-buildbot | build #460 of nightly_all is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_all/builds/460 | 03:32 |
-!- zxtx [~zv@149.165.244.2] has joined #shogun | 05:16 | |
-!- sonne|osx [~sonne@82.113.121.195] has joined #shogun | 08:20 | |
wiking | sonne|osx: yo. have u managed to update the notebook update script? | 08:20 |
sonne|osx | wiking: ? | 08:20 |
sonne|osx | so no | 08:20 |
wiking | :( | 08:20 |
sonne|osx | wiking: is this from the automated tests? | 08:21 |
sonne|osx | d | 08:21 |
sonne|osx | d/home/buildslave/nightly_none/build/tests/unit/base/SGObject_unittest.cc: In function 'void* stress_test(void*)': | 08:21 |
sonne|osx | SGObject_unittest.cc:56:17: warning: unused variable 'labs_2' [-Wunused-variable] | 08:21 |
wiking | sonne|osx: no that's something handwritten | 08:22 |
sonne|osx | ok then | 08:24 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 08:24 | |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 687d5c8 / tests/unit/base/SGObject_unittest.cc: https://github.com/shogun-toolbox/shogun/commit/687d5c83fc19e690988d261304250d76fcf03e76 | 08:24 |
shogun-notifier- | shogun: disable refcount test when refcounting is not available | 08:24 |
sonne|osx | wiking: and the Serialization* tests from here http://buildbot.shogun-toolbox.org/builders/nightly_all/builds/460/steps/test/logs/stdio ? | 08:25 |
wiking | Serialization{JSON, Ascii, XML, HDF5} are autogenerated from the jinja2 template | 08:26 |
sonne|osx | wiking: what are those doing? | 08:26 |
wiking | sonne|osx: serialize all the possible objects into a file | 08:26 |
wiking | reads it | 08:26 |
sonne|osx | wiking: we really need a gdb trace | 08:26 |
wiking | and checks if it's equal | 08:26 |
wiking | sonne|osx: couldn't find such option for ctest yet | 08:27 |
sonne|osx | I have no clue how to debug this | 08:27 |
sonne|osx | wiking: no we can do it when running our test | 08:27 |
sonne|osx | wiking: like this | 08:27 |
wiking | sonne|osx: how... do u know the ctest command for it? | 08:27 |
sonne|osx | test | 08:27 |
wiking | btw: do you know how to create a tmpfs? | 08:27 |
sonne|osx | test || gdb test | 08:27 |
sonne|osx | wiking: isn't ctest just calling some commands | 08:28 |
sonne|osx | ? | 08:28 |
wiking | afaik yes | 08:28 |
sonne|osx | so then it is easy | 08:28 |
sonne|osx | a test returns 0 if it worked fine | 08:29 |
sonne|osx | if not we can re-run the test with gdb | 08:29 |
sonne|osx | and voila! | 08:29 |
sonne|osx | this is how I did this when we had configure running | 08:30 |
sonne|osx | maybe you can do it a bit more beautiful by defining some cmake function that is encapsulating this? | 08:30 |
wiking | dunno | 08:31 |
sonne|osx | shogun-buildbot: force build branch=develop 'nightly_none' | 08:31 |
shogun-buildbot | no such builder 'branch=develop' | 08:31 |
sonne|osx | shogun-buildbot: force build --branch=develop 'nightly_none' | 08:31 |
shogun-buildbot | The build has been queued, I'll give a shout when it starts | 08:32 |
sonne|osx | wiking: so can you do it? | 08:34 |
sonne|osx | wiking: all that is needed is to do | 08:34 |
sonne|osx | in: /bin/ls || gdb --command shogun/src/.gdb /bin/ls | 08:35 |
shogun-buildbot | build #164 of osx1 - libshogun is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/osx1%20-%20libshogun/builds/164 blamelist: Soeren Sonnenburg <sonne@debian.org> | 08:35 |
sonne|osx | gtg brb | 08:35 |
-!- sonne|osx [~sonne@82.113.121.195] has quit [Quit: sonne|osx] | 08:35 | |
shogun-buildbot | build #474 forced | 08:38 |
shogun-buildbot | I'll give a shout when the build finishes | 08:38 |
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 08:38 | |
shogun-buildbot | build #474 of nightly_none is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_none/builds/474 | 08:41 |
shogun-notifier- | shogun: Viktor Gal :develop * 83b821f / src/shogun/mathematics/Mosek.cpp: https://github.com/shogun-toolbox/shogun/commit/83b821f949c07762018309025686a852b709be92 | 08:50 |
shogun-notifier- | shogun: Update Mosek.cpp wrapper to support mosek 7 | 08:50 |
shogun-notifier- | shogun: issue #1287 | 08:50 |
-!- lisitsyn [~lisitsyn@fb2-lo1.global63.net] has quit [Quit: Leaving.] | 09:06 | |
shogun-buildbot | build #1503 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1503 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:10 |
shogun-buildbot | build #1222 of rpm1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/rpm1%20-%20libshogun/builds/1222 blamelist: Viktor Gal <viktor.gal@maeth.com> | 09:17 |
sonne|work | wiking: could you please disable running of tests on the nightly_none bot? | 09:27 |
wiking | sonne|work: why? | 09:27 |
-!- travis-ci [~travis-ci@ec2-54-242-103-128.compute-1.amazonaws.com] has joined #shogun | 09:27 | |
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/11582382 | 09:27 |
-!- travis-ci [~travis-ci@ec2-54-242-103-128.compute-1.amazonaws.com] has left #shogun [] | 09:27 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting] | 09:28 | |
sonne|work | I think it is enough to see that things compile. Many examples/tests have to be written differently if certain stuff is not avilable | 09:28 |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 09:28 | |
wiking | ^ disabled | 09:28 |
sonne|work | sure all doable but what the heck is it worth the effort? | 09:28 |
sonne|work | if we figure this is a problem we can enable it later | 09:29 |
sonne|work | wiking: and can you do the test running stuff with gdb or not? | 09:31 |
wiking | sonne|work: not now | 09:31 |
wiking | and i think not in the next 3 days | 09:31 |
shogun-notifier- | shogun: Viktor Gal :develop * ac174a6 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/ac174a6b111d7f1d1457bec0e11f1c3ca8a09ea4 | 09:35 |
shogun-notifier- | shogun: Add Mosek based libshogun examples if Mosek is enabled | 09:35 |
shogun-notifier- | shogun: Fix Mosek env creation in CCSOSVM for mosek v7 | 09:35 |
shogun-notifier- | shogun: add USE_MOSEK macro to config.h | 09:35 |
sonne|work | shogun-buildbot: force build --branch=develop 'nightly_none' | 09:36 |
shogun-buildbot | The build has been queued, I'll give a shout when it starts | 09:36 |
sonne|work | wiking: how can one run e.g. piped commands in cmake? | 09:38 |
sonne|work | wiking: such as echo "foo" | sed 's/o/a' ? | 09:39 |
wiking | sometimes it works when u add two COMMAND lines | 09:41 |
wiking | say | 09:41 |
-!- lambday [67157f37@gateway/web/freenode/ip.103.21.127.55] has joined #shogun | 09:42 | |
wiking | COMMADN echo "foo" | 09:42 |
wiking | COMMAND sed ... | 09:42 |
shogun-buildbot | build #475 forced | 09:43 |
shogun-buildbot | I'll give a shout when the build finishes | 09:43 |
wiking | lambday: hey | 09:44 |
lambday | wiking: hi | 09:44 |
lambday | wiking: thanks for the SGVector | 09:45 |
wiking | lambday: could u use it? | 09:45 |
lambday | wiking: not yet tested | 09:45 |
lambday | checking | 09:45 |
lambday | :) | 09:45 |
wiking | ok | 09:46 |
wiking | let me know if something is not ok or u need something more | 09:47 |
wiking | i've tested SGVector<complex> | 09:47 |
wiking | but not the rest | 09:47 |
lambday | wiking: I need SGSparseMatrix<complex>, but I wanna try with this | 09:48 |
lambday | so I need to add this in and out for sparse in swig_typemaps.i with NPY_CDOUBLE | 09:49 |
lambday | right? | 09:49 |
lambday | anything else? | 09:49 |
shogun-buildbot | build #1504 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1504 blamelist: Viktor Gal <viktor.gal@maeth.com> | 09:50 |
wiking | lambday: i've added SGSparseMatrix<complex | 09:50 |
lambday | wiking: awesome! I am testing then :) | 09:51 |
lambday | I thought we needed to add TYPEMAP_SPARSEFEATURES_IN/OUT(complex64_t, NPY_CDOUBLE).. earlier I was trying with this one | 09:52 |
wiking | that's for SparseFeatures | 09:54 |
shogun-buildbot | build #20 of clang34 - static analysis is complete: Failure [failed analyse] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/20 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:57 |
shogun-buildbot | build #475 of nightly_none is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_none/builds/475 | 09:57 |
-!- travis-ci [~travis-ci@ec2-54-234-21-192.compute-1.amazonaws.com] has joined #shogun | 10:06 | |
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/11583062 | 10:06 |
-!- travis-ci [~travis-ci@ec2-54-234-21-192.compute-1.amazonaws.com] has left #shogun [] | 10:06 | |
shogun-buildbot | build #8 of clang34 - undefined behaviour analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/8 blamelist: Soeren Sonnenburg <sonne@debian.org>, Viktor Gal <viktor.gal@maeth.com> | 10:08 |
shogun-buildbot | build #12 of clang34 - thread analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/12 blamelist: Soeren Sonnenburg <sonne@debian.org>, Viktor Gal <viktor.gal@maeth.com> | 10:24 |
sonne|work | wiking: this is how it works http://stackoverflow.com/questions/3065220/ctest-with-multiple-commands | 10:33 |
shogun-buildbot | build #1223 of rpm1 - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/rpm1%20-%20libshogun/builds/1223 | 10:36 |
wiking | that's good | 10:37 |
wiking | needs implementing then | 10:37 |
shogun-buildbot | build #109 of precise - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/109 blamelist: Viktor Gal <viktor.gal@maeth.com> | 10:44 |
-!- hushell [~hushell@c-98-232-178-161.hsd1.or.comcast.net] has quit [Quit: WeeChat 0.3.7] | 10:48 | |
wiking | HeikoS: around? | 10:50 |
wiking | HeikoS: i need to run something on a cluster :) | 10:50 |
-!- travis-ci [~travis-ci@ec2-23-22-84-25.compute-1.amazonaws.com] has joined #shogun | 10:57 | |
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/11584074 | 10:57 |
-!- travis-ci [~travis-ci@ec2-23-22-84-25.compute-1.amazonaws.com] has left #shogun [] | 10:57 | |
shogun-buildbot | build #1505 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1505 blamelist: Viktor Gal <viktor.gal@maeth.com> | 10:59 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 11:03 | |
shogun-buildbot | build #21 of clang34 - static analysis is complete: Failure [failed analyse] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/21 blamelist: Viktor Gal <viktor.gal@maeth.com> | 11:12 |
shogun-buildbot | build #1853 of deb3 - modular_interfaces is complete: Failure [failed test ruby modular] Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1853 blamelist: Viktor Gal <viktor.gal@maeth.com> | 11:20 |
wiking | lambday: or u would need SparseFeatures<complex64_t> wrapper as well? | 11:30 |
wiking | lambday: plz comment what you've find out about the mapping (works/does work) at: https://github.com/shogun-toolbox/shogun/issues/1493 | 11:32 |
lambday | wiking: seems like I needed to add the sparse features thing after all, although it says features it actually works on sparse matrix | 11:32 |
lambday | earlier I couldn't make it work | 11:32 |
lambday | I added it | 11:32 |
lambday | and now it works | 11:32 |
lambday | SGVector<complex64_t> works perfectly by the way :) :) | 11:33 |
lambday | thanks a lot man :) | 11:33 |
wiking | ok | 11:33 |
wiking | so u need sparsefeatures as well? | 11:33 |
lambday | wiking: I will send a PR, with a few of my wrappers | 11:33 |
lambday | wiking: actually, no.. I only needed SGSparseMatrix<complex64_t> to work | 11:34 |
lambday | wiking: just a sec | 11:34 |
wiking | lambday: that should work | 11:34 |
wiking | although i have not tested it | 11:34 |
wiking | ComplexSparseMatrix | 11:34 |
lambday | wiking: https://github.com/shogun-toolbox/shogun/blob/develop/src/interfaces/python_modular/swig_typemaps.i#L1216 | 11:34 |
wiking | ? | 11:35 |
lambday | when I added TYPEMAP_SPARSEFEATURES_OUT(complex64_t, NPY_CDOUBLE), then it worked | 11:35 |
wiking | before that? | 11:35 |
lambday | didn't :( | 11:35 |
lambday | the macro is for SGSparseMatrix | 11:36 |
lambday | not CSparseFeatures | 11:36 |
lambday | :( | 11:36 |
wiking | In [1]: from modshogun import ComplexSparseMatrix | 11:36 |
wiking | In [2]: x = ComplexSparseMatrix() | 11:36 |
wiking | this works for me | 11:36 |
wiking | w/o that line | 11:36 |
wiking | make up your mind | 11:36 |
wiking | do u need SparseMatrix | 11:36 |
wiking | or SparseFeatures | 11:36 |
lambday | wiking: but I tried to pass a csc_matrix(complex) as an argument to a class which takes SGSparseMatrix<complex64_t> it didn't work | 11:37 |
lambday | I need sparse matrix | 11:37 |
wiking | csc_matrix(complex)? | 11:37 |
wiking | what's this? | 11:37 |
lambday | that's scipy.sparse matrix... | 11:38 |
lambday | for real types, I can seamlessly pass a csc_matrix to a shogun method that takes SGSparseMatrix<float64_t> | 11:38 |
lambday | for complex, I needed the same | 11:38 |
wiking | ok tell me the code that i can generate this scipy.sparse | 11:39 |
wiking | csc_matrix(complex | 11:39 |
lambday | wiking: wait.. | 11:39 |
wiking | as you are talking way out of context | 11:39 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun | 11:39 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 11:39 | |
@iglesiasg | Hi people! | 11:39 |
wiking | yo | 11:39 |
lambday | wiking: take any matrix, a=array(random.randn(10,10)) | 11:40 |
lambday | wiking: then b=csc_matrix(a) | 11:40 |
@iglesiasg | wiking: ipython notebooks ready then? :O | 11:40 |
lambday | oh and import csc_matrix from scipy.sparse | 11:40 |
wiking | lambday: tell me the imports | 11:40 |
lambday | from scipy.sparse import csc_matrix | 11:40 |
wiking | what's arra | 11:41 |
wiking | array | 11:41 |
wiking | random | 11:41 |
lambday | numpy array | 11:41 |
wiking | a=np.array(random.randn(10,10)) | 11:41 |
wiking | --------------------------------------------------------------------------- | 11:41 |
wiking | NameError Traceback (most recent call last) | 11:41 |
wiking | <ipython-input-13-1536df4033d4> in <module>() | 11:41 |
wiking | ----> 1 a=np.array(random.randn(10,10)) | 11:41 |
wiking | NameError: name 'random' is not defined | 11:41 |
wiking | ... | 11:41 |
lambday | from numpy import * | 11:41 |
wiking | iglesiasg: well test it | 11:41 |
lambday | its in numpy | 11:42 |
wiking | ok | 11:42 |
wiking | i have b now | 11:42 |
lambday | yeah | 11:42 |
wiking | what now? | 11:43 |
lambday | wiking: now what I needed was I needed to construct one of my class CSparseMatrixOperator<complex64_t> that takes a SGSparseMatrix<complex64_t> as an arg | 11:43 |
lambday | shit | 11:43 |
lambday | I added it locally you won't be able to access it | 11:43 |
lambday | what I need is that, I need to pass that csc_matrix to my class | 11:43 |
lambday | in the constructor | 11:43 |
lambday | locally it works now | 11:44 |
wiking | yeah | 11:44 |
wiking | CSparseMatrixOperator needs a mapping as well | 11:44 |
wiking | :) | 11:44 |
lambday | wiking: yeah I added that locally | 11:44 |
lambday | sending PR :) | 11:44 |
wiking | but i dont know how you can set the values of ComplexSparseMatrix | 11:44 |
wiking | :< | 11:44 |
wiking | in fact i dont know how u do that with ComplexMatrix either | 11:44 |
lambday | no idea! | 11:44 |
wiking | :P | 11:44 |
wiking | or RealMatrix | 11:45 |
wiking | how do you convert a numpy.matrix -> RealMatrix | 11:45 |
lambday | but if I pass a csc_matrix to a method that takes SGSparseMatrix as an arg, it works :D | 11:45 |
wiking | or RealSparseMatrix | 11:45 |
wiking | iglesiasg: btw i think it would be a good thing to enable director classes on that shogun install in the notebook server | 11:46 |
@iglesiasg | wiking: using the image I guess, right? | 11:46 |
lambday | wiking: I am using wrappers for these, so in the constructor I pass csc_matrix, works fine | 11:46 |
wiking | iglesiasg: image? | 11:46 |
wiking | iglesiasg: what image do you mean? | 11:46 |
@iglesiasg | wiking: to test the notebooks | 11:46 |
wiking | no comprende | 11:46 |
@iglesiasg | haha | 11:47 |
wiking | elaborate plz | 11:47 |
wiking | :) | 11:47 |
@iglesiasg | wiking: 11:41 < wiking> iglesiasg: well test it | 11:47 |
wiking | iglesiasg: ah no | 11:47 |
wiking | iglesiasg: haven't u received my email to shogun-team@ ? | 11:47 |
@iglesiasg | nope | 11:47 |
wiking | weeeeird | 11:47 |
@iglesiasg | that was the part of | 11:47 |
@iglesiasg | developers have been sent blah blah I did not get | 11:48 |
@iglesiasg | but it might have missed it | 11:48 |
@iglesiasg | let me double check... | 11:48 |
wiking | ok let me know | 11:48 |
wiking | maybe it's still in greylist | 11:48 |
wiking | oooor i've sent from an address from which i'm not allowed to that list | 11:48 |
wiking | oh ok | 11:49 |
wiking | that's what happend | 11:49 |
wiking | let me resend it | 11:49 |
wiking | iglesiasg: sent again... let me know if u received now | 11:49 |
@iglesiasg | wiking: received | 11:49 |
wiking | cool | 11:49 |
wiking | sorry about the fuckup | 11:49 |
@iglesiasg | wiking: aah np | 11:51 |
@iglesiasg | I love struggling with these little things :D | 11:51 |
wiking | ah shiiit | 11:53 |
wiking | we would need this | 11:53 |
wiking | Magics for temporary workspace | 11:53 |
wiking | https://github.com/tkf/ipython-tempmagic | 11:53 |
@iglesiasg | testing the notebooks atm... | 11:53 |
@iglesiasg | cool stuff | 11:53 |
@iglesiasg | wiking: to ensure, so every time I run a cell | 11:53 |
@iglesiasg | wiking: it is being executed in 7nn, right? | 11:54 |
wiking | yesp | 11:54 |
wiking | *yeps | 11:54 |
@iglesiasg | sweet! | 11:54 |
wiking | the problem is that currently | 11:54 |
wiking | or not the problem | 11:54 |
wiking | but the fact is | 11:54 |
@iglesiasg | it is saved | 11:54 |
wiking | that the directory where you enter (where the notebooks are) | 11:54 |
wiking | is readonly | 11:54 |
@iglesiasg | why problem? | 11:55 |
@iglesiasg | drawback? | 11:55 |
wiking | well | 11:55 |
wiking | can u add a new notebook now? | 11:55 |
wiking | 2013-09-20 09:50:58.338 [tornado.general] WARNING | 400 GET /new (91.64.72.127): Unexpected error while autosaving notebook: [Errno 30] Read-only file system: u'/home/notebook/notebooks/Untitled0.ipynb' | 11:55 |
wiking | WARNING:tornado.general:400 GET /new (91.64.72.127): Unexpected error while autosaving notebook: [Errno 30] Read-only file system: u'/home/notebook/notebooks/Untitled0.ipynb' | 11:55 |
wiking | from the logs | 11:55 |
wiking | so i suspect that u cannot add a new notebook | 11:55 |
@iglesiasg | no.. I cant | 11:55 |
wiking | see | 11:56 |
wiking | that's the thing | 11:56 |
@iglesiasg | 400: Bad Request | 11:56 |
wiking | you can run the ones that are there | 11:56 |
@iglesiasg | but hold on | 11:56 |
wiking | but you cannot add a new one | 11:56 |
@iglesiasg | how do we want to handle that? | 11:56 |
@iglesiasg | I mean | 11:56 |
wiking | well this magic shit | 11:56 |
wiking | should be able to do that | 11:56 |
wiking | creates a temp location | 11:56 |
wiking | and there you can write whatever | 11:56 |
@iglesiasg | do we want people coming in creating notebooks and keeping them in our server? | 11:56 |
wiking | but then as soon as u log out | 11:56 |
@iglesiasg | or just keeping them while they are logged in? | 11:56 |
wiking | it's cleaned up | 11:57 |
wiking | well | 11:57 |
@iglesiasg | all right | 11:57 |
wiking | it'd be interesting to keep some | 11:57 |
wiking | i would say | 11:57 |
@iglesiasg | that makes sense then | 11:57 |
@iglesiasg | do you think so? | 11:57 |
wiking | maybe somebody does some interesting stuff | 11:57 |
wiking | but then again | 11:57 |
wiking | that person can send it to us | 11:57 |
@iglesiasg | mmm yeah it could be | 11:57 |
wiking | since one can download the notebooks | 11:57 |
@iglesiasg | we could train a classifier to keep the interesting ones :) | 11:57 |
wiking | hehehe | 11:57 |
wiking | but anyhow i think 2013-09-20 09:50:58.338 [tornado.general] WARNING | 400 GET /new (91.64.72.127): Unexpected error while autosaving notebook: [Errno 30] Read-only file system: u'/home/notebook/notebooks/Untitled0.ipynb' | 11:57 |
wiking | so | 11:58 |
wiking | i think that we should try to use this magic shit | 11:58 |
wiking | but yeah test it and then send some feedbacks (this is not for you but *all*) and then i'll do the changes stilll | 11:58 |
@iglesiasg | well I just realized that I need to update my notebook... | 11:59 |
@iglesiasg | we deffs need to get custom data to the notebook somehow | 11:59 |
@iglesiasg | wiking: btw, the first mail ipython notebook test mail you sent just arrived here xD | 12:00 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 12:35 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 12:38 | |
shogun-notifier- | shogun-data: hushell :master * 62b19b7 / ocr/ocr_taskar.mat: https://github.com/shogun-toolbox/shogun-data/commit/62b19b767d7e50d8e7d87466b61ba2015bd11db3 | 12:38 |
shogun-notifier- | shogun-data: added an OCR dataset (http://www.seas.upenn.edu/~taskar/ocr/) | 12:38 |
shogun-notifier- | shogun-data: Fernando Iglesias :master * a41ea33 / ocr/ocr_taskar.mat: https://github.com/shogun-toolbox/shogun-data/commit/a41ea33ee80ad01b8c690a09d98eebe374865262 | 12:38 |
shogun-notifier- | shogun-data: Merge pull request #23 from hushell/master | 12:38 |
shogun-notifier- | shogun-data: | 12:38 |
shogun-notifier- | shogun-data: added an OCR dataset (.mat) | 12:38 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Quit: Lost terminal] | 12:55 | |
lambday | HeikoS: there? | 12:56 |
lambday | wiking: sent the PR | 12:58 |
lambday | using this in my notebook now | 12:58 |
lambday | HeikoS: please have a look - http://nbviewer.ipython.org/6591994 | 12:58 |
lambday | HeikoS: will be adding ozone | 12:59 |
-!- iglesias [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun | 13:21 | |
* besser82 just runs some experimental SHOGUN-stuff with ViennaCL && CUDA and is heavily impressed! | 13:42 | |
shogun-notifier- | shogun: Fernando Iglesias :develop * 8d498b2 / examples/undocumented/libshogun/ (2 files): https://github.com/shogun-toolbox/shogun/commit/8d498b2154d0b490df6ca16ec406381e429bc3e3 | 13:47 |
shogun-notifier- | shogun: Rename and update to new API HM-SVM Mosek libshogun example | 13:47 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 3d6e5f6 / examples/undocumented/libshogun/ (2 files): https://github.com/shogun-toolbox/shogun/commit/3d6e5f6bea4619dd2ec617ab7d0e3a2484783155 | 13:47 |
shogun-notifier- | shogun: Merge pull request #1638 from iglesias/fix/hmsvm_mosek_example | 13:47 |
shogun-notifier- | shogun: | 13:47 |
shogun-notifier- | shogun: Rename and update to new API HM-SVM Mosek libshogun example | 13:47 |
-!- iglesias [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 13:50 | |
besser82 | eigenvectors do get _REALLY_ speedy with that :) | 13:53 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun | 13:57 | |
iglesiasg | sonne|work: hi! | 14:02 |
sonne|work | hi | 14:02 |
iglesiasg | have you checked the notebooks? | 14:02 |
sonne|work | a bit | 14:02 |
iglesiasg | wiking did some stuff and we can already execute them | 14:02 |
iglesiasg | cool | 14:02 |
iglesiasg | sonne|work: any thought regarding getting data into the notebook? | 14:03 |
iglesiasg | I think we really need to provide that | 14:03 |
iglesiasg | the API could look like | 14:04 |
iglesiasg | features = RealFeatures(LocalFile('path/to/file/in/client/machine')) | 14:04 |
iglesiasg | and the same for the labels | 14:05 |
sonne|work | iglesiasg: we have that already | 14:05 |
iglesiasg | oh really? | 14:05 |
sonne|work | iglesiasg: see examples RealFeatures(CSVFile('/path/to/file')) | 14:05 |
sonne|work | same with .h5 | 14:05 |
sonne|work | and libsvm files | 14:05 |
iglesiasg | sonne|work: yeah yeah | 14:05 |
sonne|work | etc | 14:05 |
iglesiasg | but the ipython runs in 7nn! | 14:05 |
iglesiasg | so CSVFile('path/blahblah') | 14:06 |
iglesiasg | would look in 7nn's FS | 14:06 |
sonne|work | sure | 14:06 |
iglesiasg | what I am thinking about is that someone using the notebook | 14:06 |
iglesiasg | can choose what data to use from her machine | 14:06 |
sonne|work | and then? | 14:07 |
iglesiasg | ? | 14:08 |
iglesiasg | well then she can do stuff in the notebook with that data | 14:08 |
sonne|work | what does he do if he has chosen some data | 14:08 |
sonne|work | that doesn't explain how to get the data | 14:08 |
sonne|work | to fatbot | 14:08 |
iglesiasg | yeah that's why I said API :) | 14:09 |
sonne|work | no idea what you say :/ | 14:09 |
sonne|work | what does features = RealFeatures(LocalFile('path/to/file/in/client/machine')) do | 14:10 |
sonne|work | where is this run? | 14:10 |
iglesiasg | LocalFile does stuff under the hood | 14:11 |
iglesiasg | sends the file | 14:11 |
iglesiasg | makes security checks | 14:11 |
iglesiasg | cheks for format | 14:11 |
iglesiasg | and finally load it in numpy array if everthing went fine | 14:12 |
sonne|work | iglesiasg: well this is maybe 2 lines python then | 14:13 |
iglesiasg | sonne|work: even better then | 14:14 |
sonne|work | import urllib2 | 14:14 |
sonne|work | req = urllib2.urlopen(url) | 14:14 |
sonne|work | content = req.read() | 14:14 |
shogun-buildbot | build #110 of precise - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/110 | 14:14 |
iglesiasg | sonne|work: what do you put in the url there? | 14:15 |
sonne|work | http://somewhere.where/the/file/is | 14:15 |
iglesiasg | I guess that file:/// ... won't work | 14:16 |
iglesiasg | because that referes to 7nn again | 14:16 |
iglesiasg | that would require the client to upload the data to a server | 14:16 |
shogun-buildbot | build #1854 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1854 | 14:24 |
shogun-buildbot | build #1855 of deb3 - modular_interfaces is complete: Failure [failed test libshogun] Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1855 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 14:25 |
-!- lambday [67157f37@gateway/web/freenode/ip.103.21.127.55] has quit [] | 14:25 | |
shogun-buildbot | build #1506 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1506 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 14:30 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 14:46 | |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun | 14:55 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 14:55 | |
shogun-buildbot | build #22 of clang34 - static analysis is complete: Failure [failed analyse] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/22 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 15:08 |
shogun-buildbot | build #1507 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1507 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 15:11 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has left #shogun [] | 15:11 | |
shogun-buildbot | build #9 of clang34 - undefined behaviour analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/9 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 15:19 |
@HeikoS | wiking: nice! just tried the could shogun :) aweseom! | 15:24 |
@iglesiasg | HeikoS: so in the example evaluation_cross_validation_classification.py | 15:30 |
@iglesiasg | HeikoS: it says that StratifiedCrossValidation is better | 15:31 |
@HeikoS | yes | 15:31 |
@iglesiasg | but standard StratifiedCrossValidatoinSplitting is also available | 15:31 |
@iglesiasg | I don't quite a get it | 15:31 |
@HeikoS | you mean CrossValidationSplitting | 15:31 |
@iglesiasg | ok | 15:31 |
@HeikoS | sometimes one cannot do stratified | 15:31 |
@iglesiasg | then you meant that :P | 15:31 |
@HeikoS | for example in regression | 15:31 |
@iglesiasg | I will update the example | 15:31 |
@HeikoS | stratified should always be used with classification | 15:31 |
@HeikoS | actually, we could think about doing something similar to stratified for regression, (grouping y values) but that would be later | 15:32 |
@HeikoS | iglesiasg: cool thanks! | 15:32 |
@iglesiasg | oh I see | 15:32 |
@iglesiasg | I guess this regression thing is related (though not limited) to GPs, right? | 15:32 |
shogun-buildbot | build #13 of clang34 - thread analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/13 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 15:33 |
@iglesiasg | HeikoS: so stratified basically maintains the same ratio of classes in the folds, right? | 15:34 |
-!- travis-ci [~travis-ci@ec2-23-22-50-150.compute-1.amazonaws.com] has joined #shogun | 15:39 | |
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/11590331 | 15:39 |
-!- travis-ci [~travis-ci@ec2-23-22-50-150.compute-1.amazonaws.com] has left #shogun [] | 15:39 | |
@HeikoS | iglesiasg: yes | 15:45 |
@HeikoS | iglesiasg: this reduces the variance of the test error estimate | 15:45 |
@HeikoS | quite a bit | 15:45 |
@iglesiasg | HeikoS: aham, stupid question: the ratio wrt to the whole data set, I guess | 15:46 |
@HeikoS | iglesiasg: yeah as far as this is possible | 15:46 |
@HeikoS | iglesiasg: it just distributes the labels to the folds in turns | 15:46 |
@HeikoS | giving one class to a fold a t a time | 15:47 |
-!- lambday [67157f37@gateway/web/freenode/ip.103.21.127.55] has joined #shogun | 15:47 | |
@iglesiasg | oh I understand | 15:49 |
@iglesiasg | it is a simple idea, like the great ones! hehe | 15:50 |
@HeikoS | iglesiasg: there is a nice paper about these things | 15:50 |
@HeikoS | I think its mentioend in the source | 15:50 |
-!- zxtx [~zv@149.165.244.2] has quit [Ping timeout: 245 seconds] | 16:00 | |
lambday | HeikoS: checking your mail :) | 16:08 |
@HeikoS | lambday: I kind of wrote a lot of things, but most of them are small | 16:08 |
@HeikoS | lambday: main thing you should work on is to make it flow a bit more and look a bit nicer, but its already very cool | 16:09 |
@HeikoS | lambday: and its a lot of stuff, thats why I would put the most interesting parts with the prettiest pictures on top :) | 16:09 |
-!- lambday_ [67157e37@gateway/web/freenode/ip.103.21.126.55] has joined #shogun | 16:10 | |
lambday_ | HeikoS: sorry I got dc | 16:11 |
@HeikoS | lambdayno worries | 16:11 |
-!- lambday [67157f37@gateway/web/freenode/ip.103.21.127.55] has quit [Ping timeout: 250 seconds] | 16:12 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 16:47 | |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun | 17:26 | |
-!- FSCV [~FSCV@65.19.131.247] has joined #shogun | 17:28 | |
-!- lambday_ [67157e37@gateway/web/freenode/ip.103.21.126.55] has quit [] | 17:32 | |
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has joined #shogun | 17:35 | |
-!- FSCV_ [~FSCV@187.191.50.146] has joined #shogun | 17:46 | |
-!- FSCV [~FSCV@65.19.131.247] has quit [Ping timeout: 264 seconds] | 17:48 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 18:12 | |
shogun-notifier- | shogun: Heiko Strathmann :develop * c82e050 / src/shogun/mathematics/linalg/ (4 files): https://github.com/shogun-toolbox/shogun/commit/c82e05033eca8e96a625c60dd020ab2291aa842c | 18:12 |
shogun-notifier- | shogun: turned some debug messages into info messages | 18:12 |
shogun-notifier- | shogun: Heiko Strathmann :develop * da19157 / src/shogun/mathematics/linalg/ (4 files): https://github.com/shogun-toolbox/shogun/commit/da191575dad9865be95ea51fbbd8c2c3524dbce9 | 18:12 |
shogun-notifier- | shogun: Merge pull request #1641 from karlnapf/develop | 18:12 |
shogun-notifier- | shogun: | 18:12 |
shogun-notifier- | shogun: turned some debug messages into info messages | 18:12 |
@HeikoS | iglesiasg: https://travis-ci.org/shogun-toolbox/shogun/jobs/11587959 | 18:13 |
@HeikoS | lmnn still fails | 18:13 |
shogun-notifier- | shogun: lambday :develop * e39fa9e / src/interfaces/ (3 files): https://github.com/shogun-toolbox/shogun/commit/e39fa9ed85d4b310a407472b345fc80049dfb7d4 | 18:14 |
shogun-notifier- | shogun: added ComplexSparseMatrixOperator to modular interface (log-det) | 18:14 |
shogun-notifier- | shogun: Heiko Strathmann :develop * c6e84fd / src/interfaces/ (3 files): https://github.com/shogun-toolbox/shogun/commit/c6e84fd1c432cc30f30d9b65465f0cc3a070f42e | 18:14 |
shogun-notifier- | shogun: Merge pull request #1637 from lambday/feature/log_determinant | 18:14 |
shogun-notifier- | shogun: | 18:14 |
shogun-notifier- | shogun: added ComplexSparseMatrixOperator to modular interface (log-det) | 18:14 |
@HeikoS | votjakovr: looking at the PR now | 18:14 |
votjakovr | HeikoS: thanks | 18:15 |
@HeikoS | votjakovr: so whats the status? any questions on the PR? when is it ready to merge? what is missing? | 18:16 |
votjakovr | HeikoS: i've used some trick there to make it possible to call SG_ADD with multiply arguments | 18:16 |
@HeikoS | votjakovr: very nice! we could even do this for MS_NOT_AVAILABLE then | 18:16 |
@HeikoS | so that one only has to add it if modelselection is available | 18:17 |
@HeikoS | but no need to change that now | 18:17 |
@HeikoS | votjakovr: pls dont do all these whitespace changes, makes it hard to look through PRs | 18:17 |
votjakovr | HeikoS: yep, i know, sorry | 18:17 |
@iglesiasg | HeikoS: will take a look at LMNN again | 18:18 |
@iglesiasg | can it be that it was for a commit where develop was not rebased? | 18:19 |
@iglesiasg | gtg now | 18:19 |
@iglesiasg | but will check later | 18:19 |
@HeikoS | iglesiasg: I dont know, but then this is on travis since its not possible to merge otherwise | 18:20 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has quit [Quit: Lost terminal] | 18:22 | |
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has quit [Quit: Leaving.] | 18:32 | |
@HeikoS | votjakovr: what is missing now? | 18:44 |
@HeikoS | votjakovr: I will be away on the weekend, can maybe help a bit with the notebook on late Sat or Sun, but I need your initial draft for that | 18:45 |
@HeikoS | votjakovr: its impossible for me to read through all of these changes, especially since there so many unreleated changes, so what are you working on currently? | 18:45 |
votjakovr | HeikoS: i'm still working on CParameterCombination | 18:48 |
@HeikoS | votjakovr: whats after that? | 18:48 |
votjakovr | HeikoS: gradient model selection will be working | 18:48 |
votjakovr | HeikoS: then notebook | 18:49 |
@HeikoS | votjakovr: okay nice, I just skipped though the PR, seems al ok | 18:49 |
@HeikoS | votjakovr: but the problem is I cannot read this properly since my browser can't handle such a large page ;) | 18:49 |
shogun-buildbot | build #1856 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1856 | 18:49 |
@HeikoS | votjakovr: so let me know when the thing is final and all tests pass ( I will have alook into the tests then) and then OK for merge | 18:50 |
@HeikoS | I will be around for another 2hrs | 18:50 |
@HeikoS | votjakovr: what the time schedule you have? | 18:50 |
@HeikoS | votjakovr: actually travis passed | 18:50 |
votjakovr | HeikoS: cool, i'd like to work until model selection is done | 18:51 |
@HeikoS | okay | 18:52 |
@HeikoS | votjakovr: so I will update your notebook a bit on Sun then | 18:52 |
@HeikoS | votjakovr: let me know how the state is in 1-2 hrs ok? | 18:52 |
votjakovr | HeikoS: ok, about examples: i drop them, because they are using old gradient model selection | 18:53 |
votjakovr | HeikoS: it's not a problem to back them | 18:53 |
@HeikoS | votjakovr: I see, then pls rather deactivate them for now | 18:53 |
@HeikoS | since all these method need to be illustrated atsome point | 18:54 |
@HeikoS | just do something like #ifdef 0 | 18:54 |
@HeikoS | and add a comment that they are currently disabled due to api changes but soon coming again | 18:54 |
votjakovr | HeikoS: ok, i'll do | 18:54 |
shogun-buildbot | build #1508 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1508 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 18:55 |
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has joined #shogun | 18:58 | |
lambday | HeikoS: sorry I had to go because had a meeting with prof' | 18:59 |
lambday | HeikoS: I have some doubts regarding your suggestions.. mailing it | 19:00 |
shogun-buildbot | build #1857 of deb3 - modular_interfaces is complete: Failure [failed test ruby modular] Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1857 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 19:09 |
@HeikoS | lambday: ok let me know | 19:09 |
@HeikoS | nothing there yet | 19:09 |
lambday | HeikoS: just a few things.. incorporating the rest of the things | 19:13 |
@HeikoS | lambday: okay, keep in mind that these are just suggestions :) | 19:13 |
lambday | HeikoS: yeah but this will make it look cool :) | 19:13 |
lambday | HeikoS: how do I display the cholesky factors? using cholmod? | 19:14 |
lambday | using shogun I cannot do that | 19:14 |
@HeikoS | lambday: turn the matrix into dense and compute dense cholesky | 19:15 |
@HeikoS | for the small one, that should work | 19:15 |
lambday | HeikoS: so, for west0479 I can do that | 19:16 |
lambday | and you want the putting it all together part pulled above | 19:16 |
@HeikoS | lambday: yeah exactly | 19:16 |
@HeikoS | lambday: yeah I think so | 19:16 |
@HeikoS | all these solvers are not crucial for using the method | 19:16 |
lambday | nah | 19:16 |
@HeikoS | more like an illustration how to use the tools you created | 19:17 |
lambday | yes | 19:17 |
lambday | alright.. | 19:17 |
@HeikoS | you can introduce that as "All the methods needed to estimate the log-determinants in the above way can also be used on their own | 19:17 |
lambday | ah that sounds great' | 19:17 |
lambday | HeikoS: oh one thing.. I didn't get that part when you said that the matrix I used for CG may not be psd | 19:18 |
lambday | I already did m^Tm+I | 19:18 |
lambday | and applied cg on that one | 19:18 |
@HeikoS | ah | 19:18 |
@HeikoS | maybe I missed that | 19:18 |
@HeikoS | sorry then | 19:18 |
lambday | where m is randomly generated | 19:18 |
lambday | HeikoS: np.. oh and where would these notebooks go? (so that I can put a relative path to the data, gotta add apache2 and ozone though) | 19:19 |
@HeikoS | there is a directory for them ... | 19:19 |
@HeikoS | in doc | 19:20 |
lambday | HeikoS: doc/ipython-notebooks | 19:20 |
lambday | found it | 19:20 |
lambday | LD | 19:20 |
lambday | :D | 19:20 |
@HeikoS | lambday: just respect the directory structure for yours | 19:20 |
lambday | alright | 19:21 |
lambday | HeikoS: I wanted to add something for the ozone mcmc | 19:21 |
lambday | HeikoS: oh and one more thing... combining the normal and probing samplers is exactly what I did earlier | 19:21 |
lambday | HeikoS: but probing samplers have so less variance,that when I put it with Normal, it almost looks aligned with the actual logdet | 19:22 |
@HeikoS | lambday: I see | 19:22 |
lambday | that's why kept these two separate | 19:22 |
lambday | :-/ | 19:22 |
@HeikoS | then do subplot(121) #plot normal subplot(122) #plot probing | 19:22 |
@HeikoS | just do have them next to each other | 19:22 |
@HeikoS | and plot histograms on top of each other | 19:23 |
@HeikoS | would be cool to have a histogram like on the poster | 19:23 |
@HeikoS | with both samples and the true value | 19:23 |
lambday | HeikoS: alright | 19:23 |
lambday | still new to this matplotlib, fighting a bit :-/ | 19:23 |
lambday | HeikoS: can I add external images (that are not generated plots from the code) in the notebook? | 19:24 |
@HeikoS | lambday: yes, but why do you want to do that? | 19:24 |
@HeikoS | lambday: ask me if you have problems with plotting | 19:24 |
lambday | HeikoS: I wanted to put a CG-M convergence plot based on the residuals | 19:24 |
@HeikoS | lambday: can you actually access the residuals? | 19:25 |
@HeikoS | lambday: thats a nice idea in fact | 19:25 |
@HeikoS | Cant you do that for a small matrix, extract the residuals and then just plot? | 19:25 |
lambday | HeikoS: externally, run the c++ in debug more, collect the residuals (thanks to you) | 19:25 |
lambday | HeikoS: how can I externally insert an image to a notebook.. it works locally, but while uploading it in the gist to the nbviewer, I don't find an option to upload any image | 19:26 |
@HeikoS | lambday: I dont like this so much with the external picture | 19:27 |
@HeikoS | lambday: though I like the idea about plotting convergence | 19:27 |
lambday | HeikoS: then I have to add something in the solvers itself to store the residuals :-/ | 19:27 |
@HeikoS | lambday: yeah, in fact that is useful | 19:28 |
@HeikoS | lambday: but it should be off by default | 19:28 |
lambday | HeikoS: is it :-/ | 19:28 |
lambday | alright | 19:28 |
@HeikoS | lambday: yes, in cases one is interested in acessing residuals | 19:28 |
besser82 | HeikoS, wiking, sonne|work: Anyone of you interested in adding ViennaCL-support to SHOGUN? | 19:28 |
@HeikoS | lambday: can this be done once? and then used form all solvers? | 19:28 |
lambday | HeikoS: alright then | 19:28 |
@HeikoS | lambday: but maybe do other things first | 19:28 |
@HeikoS | lambday: its not too important | 19:28 |
@HeikoS | lambday: but the external picture is not really useful if there is not code how to generate it | 19:29 |
lambday | HeikoS: could you please explain a bit regarding the ozone model | 19:29 |
@HeikoS | and since convergence diagnostics are always useful, should be in there | 19:29 |
lambday | HeikoS: yeah I understand | 19:29 |
@HeikoS | lambday: ok, what to explain? | 19:29 |
lambday | HeikoS: I didn't get what these matrices represent? | 19:29 |
lambday | GiCG, C0, G A | 19:29 |
@HeikoS | besser82: maybe at some point :) | 19:30 |
@HeikoS | besser82: would be cool to have these kind of backend | 19:30 |
@HeikoS | besser82: though we are currently heavily using eigen3 | 19:30 |
@HeikoS | lambday: ah | 19:30 |
@HeikoS | lambday: thats not too important, it is the model | 19:30 |
@HeikoS | so A is a piecewise linear triangulation of a sphere (the globe) | 19:30 |
lambday | (with Q, we are not even using A, right?) | 19:31 |
@HeikoS | it takes the latent gaussian field of the model (which lives on a sphere) and maps it to a plane | 19:31 |
@HeikoS | lambday: there are two log-dets in the log likelihood | 19:31 |
@HeikoS | once involves A^T A | 19:31 |
besser82 | HeikoS, wiking, sonne|work: GTX285-SLI would slightly outperform recent Xeons, but on affordable $. :) | 19:31 |
-!- FSCV_ [~FSCV@187.191.50.146] has quit [Quit: Leaving] | 19:31 | |
besser82 | HeikoS: eigenvector runs really speedy on ViennaCL. Just tried :) | 19:32 |
@HeikoS | besser82: thats really nice, but the problem is all the interfacing | 19:32 |
@HeikoS | besser82: we dont want to change everything again, just moved to eigen3 | 19:32 |
@HeikoS | if eigen3 had a backend like that, that would be coo | 19:32 |
@HeikoS | cool | 19:32 |
@HeikoS | and is rather the way to go | 19:33 |
@HeikoS | ah I just see there is eigen stuff in there | 19:33 |
besser82 | HeikoS: wouldn't be much for eigen3 --> ViennaCL. ViennaCL has interface for eigen3 :) | 19:33 |
@HeikoS | besser82: so that might be an idea | 19:33 |
@HeikoS | yep, that sounds cool | 19:33 |
-!- van51 [~van51@athedsl-409794.home.otenet.gr] has joined #shogun | 19:33 | |
@HeikoS | besser82: maybe a gsoc project or so :) | 19:33 |
lambday | what's ViennaCL? | 19:34 |
@HeikoS | lambday: dont worry for now :) but here http://viennacl.sourceforge.net/viennacl-about.html | 19:34 |
besser82 | lambday: ViennaCL is BLAS, LAPACK/ATLAS, eigen3 on OpenCL/CUDA :) | 19:34 |
lambday | HeikoS: just curious :D | 19:34 |
@HeikoS | lambday: so you could add a little section before or after the tools section | 19:34 |
@HeikoS | which states the ozone model | 19:35 |
lambday | HeikoS: tools? | 19:35 |
@HeikoS | solvers etc | 19:35 |
@HeikoS | and kind of motivates why we need the log-determinants | 19:35 |
@HeikoS | and then runs the thing with very low max iterations | 19:35 |
@HeikoS | to illustrate that it works on desktop computers | 19:35 |
lambday | HeikoS: why we need log-determinants --- is coming from the requirement of computing the log-likelihood itself, right? | 19:36 |
lambday | HeikoS: sorry I am not getting what you mean here :( | 19:36 |
-!- sonne|osx [~sonne@f053037117.adsl.alicedsl.de] has joined #shogun | 19:36 | |
@HeikoS | lambday: exactly, every GMRF needs log.determinants | 19:36 |
@HeikoS | in fact every Gaussian distribution | 19:36 |
shogun-buildbot | build #1509 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1509 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com> | 19:36 |
shogun-buildbot | build #23 of clang34 - static analysis is complete: Failure [failed analyse] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/23 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 19:37 |
lambday | HeikoS: yes | 19:37 |
lambday | darn you cyg1 :'( | 19:37 |
@HeikoS | but the precision matrix is sparse in every GMRF | 19:37 |
@HeikoS | so thats just an example | 19:37 |
@HeikoS | you can point to the Aune et al paper | 19:37 |
@HeikoS | or even to this one | 19:37 |
lambday | besser82: sounds really interesting | 19:38 |
lambday | :D | 19:38 |
lambday | besser82: I have no clue regarding the last two OpenCL/.CUDA :D | 19:38 |
lambday | HeikoS: :-/ | 19:38 |
@HeikoS | http://arxiv.org/pdf/1104.3436.pdf | 19:39 |
@HeikoS | page 22 bottom | 19:39 |
@HeikoS | you could reference this | 19:39 |
besser82 | lambday: it surely is. openCL / CUDA is distributed computing using your GPU for performing. | 19:39 |
lambday | HeikoS: checking | 19:39 |
besser82 | lambday: CUDA is sth. specific to nvidia-cards | 19:39 |
@HeikoS | besser82: this stuff is only interesting if existing libraries do this in the back | 19:39 |
@HeikoS | besser82: otherwise its way too technical to use | 19:40 |
@HeikoS | besser82: i played with GPUs a bit a few years ago | 19:40 |
@HeikoS | and without std libs this is pure horror ;= | 19:40 |
besser82 | HeikoS: ViennaCL should give simple interfaces to use for sci-tech tasks :) | 19:40 |
@HeikoS | besser82: so if we can keep our eigen3 calls as they are and just do some magic in the background, I am really up for this! | 19:41 |
besser82 | HeikoS: http://viennacl.sourceforge.net/viennacl-examples-eigen.html | 19:41 |
besser82 | HeikoS: nothing much to change :) | 19:41 |
lambday | HeikoS: I need to study :-/ what are these parameters :-/ | 19:41 |
lambday | HeikoS: I thought giving only your paper as reference for ozone | 19:42 |
@HeikoS | lambday: so, what I suggest, is just to state the model (as in the poster), mention the log-likelihood, and then show that one can compute the likelihood using shogun | 19:42 |
@HeikoS | lambday: https://github.com/karlnapf/ozone-roulette/blob/master/ozone/distribution/OzonePosterior.py | 19:43 |
@HeikoS | just take the log_likelihood method and use log_det_estimate_shogun and solve_sparse_linear_system_shogun | 19:43 |
lambday | HeikoS: and the outputs/images? | 19:44 |
lambday | oh okay | 19:44 |
@HeikoS | and say that for illustration, you only estimate the log-det but solve the linear system with cholesky | 19:44 |
besser82 | HeikoS: so just copying in/out eigen3-datatypes and performing tasks on them. | 19:44 |
@HeikoS | lambday: images would be nice | 19:44 |
@HeikoS | lambday: the plotting code is in matlab though, let me check how we can get the map | 19:44 |
lambday | HeikoS: no no.. I only want to put that much part that I myself understand :( | 19:46 |
@HeikoS | lambday: ok then, just state the model, it has an easy form since it is a std model, it just looks technical | 19:47 |
@HeikoS | so if you tell the distributions of x, y|y, tau, kappa, and then state the log-likelihood that is ok | 19:47 |
@HeikoS | and then you can plot y if you want | 19:47 |
@HeikoS | lambday: just if you want of course | 19:47 |
@HeikoS | lambday: ill send you code to plot the observations of the ozone model | 19:47 |
lambday | HeikoS: ah thanks.. :) | 19:48 |
@HeikoS | lambday: the idea here is that you just say: "Here is a sparse Gaussian model for which this stuff is useful" | 19:48 |
lambday | yes | 19:48 |
@HeikoS | and the likelihood function in the file I sent you is just directly turning the log-likelihood math into numbers, no tricks there | 19:48 |
shogun-buildbot | build #10 of clang34 - undefined behaviour analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/10 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 19:48 |
shogun-buildbot | build #1858 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/deb3%20-%20modular_interfaces/builds/1858 | 19:49 |
lambday | HeikoS: the log_likelihood method in the OzonePosterior? | 19:50 |
@HeikoS | yes | 19:50 |
@HeikoS | lambday: check it, it just computes the log-likelihood on the poster | 19:50 |
@HeikoS | lambday: I have not plotting code in python and its non-trivial to plot | 19:54 |
@HeikoS | lambday: thinking on what to do .... | 19:54 |
lambday | HeikoS: let's chuck it | 19:54 |
lambday | HeikoS: I'll add these things later on this notebook when I'll work on more on this.. | 19:54 |
@HeikoS | lambday: okay | 19:55 |
lambday | HeikoS: I really liked the issues that you posted, parallel computation engine and one other | 19:55 |
lambday | HeikoS: I would like to work on that | 19:55 |
@HeikoS | lambday: yeah we can work on that | 19:55 |
lambday | HeikoS: plus there are many many tricks that the paper suggested, I'd like to explore these too | 19:55 |
shogun-buildbot | build #165 of osx1 - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/osx1%20-%20libshogun/builds/165 | 19:55 |
@HeikoS | lambday: but I think it still would be nice to say: Here is a std GMRF model and its log likelihood. Here we load some ozone data, see reference paper, and here we can compute the log-likelihood | 19:56 |
@HeikoS | lambday: its ultra cool even without images | 19:56 |
@HeikoS | so I would totally suggest to add that | 19:56 |
lambday | HeikoS: so, computing the likelihood thing, I can use your code? :D | 19:56 |
@HeikoS | lambday: yeah sure | 19:56 |
@HeikoS | its GPL :) | 19:57 |
lambday | hehe | 19:57 |
lambday | alright | 19:57 |
lambday | yeah it would be really cool | 19:57 |
lambday | so, the structure would be ---intro --- log-det on apache2, walking through the components --- ozone --reference | 19:58 |
lambday | not just apache2, but also west0479 | 19:58 |
lambday | with fill-in | 19:58 |
lambday | I really want to put apache2 fill in cause as per source that shows worst fill-in | 19:59 |
shogun-buildbot | build #17 of debian wheezy - memcheck is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/17 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com> | 20:02 |
shogun-buildbot | build #14 of clang34 - thread analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/14 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 20:03 |
@HeikoS | lambday: okay! | 20:03 |
@HeikoS | maybe even first ozone and then components | 20:03 |
@HeikoS | lambday: but as you like | 20:03 |
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has quit [Ping timeout: 250 seconds] | 20:03 | |
@HeikoS | lambday: I have to go now, will be in touch on Sunday | 20:03 |
@HeikoS | votjakovr: I am going home now, checking later today, pls send me an email with the status today when you stop working. Good luck! | 20:06 |
-!- HeikoS [~heiko@nat-163-62.internal.eduroam.ucl.ac.uk] has left #shogun [] | 20:06 | |
-!- zxtx [~zv@149.165.244.2] has joined #shogun | 20:21 | |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has quit [Quit: Good bye!] | 20:24 | |
shogun-buildbot | build #24 of clang34 - static analysis is complete: Failure [failed analyse] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/24 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com> | 20:42 |
shogun-buildbot | build #11 of clang34 - undefined behaviour analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/11 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com> | 20:52 |
-!- zxtx [~zv@149.165.244.2] has quit [Ping timeout: 245 seconds] | 20:53 | |
shogun-buildbot | build #15 of clang34 - thread analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/15 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, lambday <heavensdevil6909@gmail.com> | 21:05 |
-!- lisitsyn [~lisitsyn@109-226-79-195.clients.tlt.100megabit.ru] has joined #shogun | 21:09 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 21:14 | |
-!- travis-ci [~travis-ci@ec2-23-23-31-73.compute-1.amazonaws.com] has joined #shogun | 21:33 | |
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/11600255 | 21:33 |
-!- travis-ci [~travis-ci@ec2-23-23-31-73.compute-1.amazonaws.com] has left #shogun [] | 21:33 | |
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:cc71:7583:fe5b:b7d1] has joined #shogun | 21:37 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 21:37 | |
@iglesiasg | sonney2k, lisitsyn ping ping | 21:39 |
lisitsyn | ja | 21:39 |
lisitsyn | iglesiasg: hello | 21:39 |
@iglesiasg | lisitsyn, how are you? | 21:39 |
lisitsyn | iglesiasg: fine | 21:39 |
lisitsyn | iglesiasg: what about you? | 21:39 |
@iglesiasg | LMNN is giving headache :D | 21:39 |
lisitsyn | iglesiasg: what's with it? | 21:39 |
@iglesiasg | https://travis-ci.org/shogun-toolbox/shogun/jobs/11599254#L4772 | 21:39 |
@iglesiasg | integration test | 21:40 |
@iglesiasg | the 16th decimal or so is different :D | 21:40 |
lisitsyn | hah | 21:40 |
@iglesiasg | lisitsyn, can that be Eigen related? | 21:41 |
lisitsyn | iglesiasg: no clue :) | 21:41 |
lisitsyn | iglesiasg: is that ok accuracy? | 21:42 |
lisitsyn | I mean 16th sign is a bit too much | 21:42 |
lisitsyn | but who knows | 21:43 |
@iglesiasg | lisitsyn, a bit too much? | 21:43 |
@iglesiasg | I think the difference is insignificant | 21:43 |
lisitsyn | iglesiasg: looks too precise | 21:43 |
@iglesiasg | lisitsyn, I have just run it more than 100 times here in my machine and it is always OK. | 21:49 |
@iglesiasg | lisitsyn, I guess I could just deactivate de integration test | 21:49 |
lisitsyn | iglesiasg: I am not sure what to do about it | 21:50 |
@iglesiasg | some of the parts are already tested in unit tests | 21:50 |
@iglesiasg | it is fine in the buildbot as well | 21:51 |
sonne|osx | lisitsyn: please disable integration tests on travis | 21:53 |
sonne|osx | it is sufficient to have this on the buildbots | 21:54 |
lisitsyn | sonne|osx: | 21:54 |
lisitsyn | ok | 21:54 |
@iglesiasg | I like that initiative! | 21:54 |
@iglesiasg | thank you guys | 21:54 |
lisitsyn | I like to disable/delete | 21:54 |
lisitsyn | :D | 21:54 |
sonne|osx | lisitsyn: can you do the gdb thing? | 21:58 |
lisitsyn | sonne|osx: yes | 21:58 |
sonne|osx | lisitsyn: https://github.com/shogun-toolbox/shogun/issues/1643 | 21:58 |
lisitsyn | I am on it just after integration tests | 21:58 |
sonne|osx | this would help *a lot* to fix e.g. cyg1 | 21:58 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 22:00 | |
shogun-notifier- | shogun: Sergey Lisitsyn :develop * 1564b47 / .travis.yml,CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/1564b47cc11cc5c1a3c6a9de6a8027a0522dcc23 | 22:00 |
shogun-notifier- | shogun: Disabled integration tests on travis | 22:00 |
@iglesiasg | lisitsyn, thanks! | 22:06 |
lisitsyn | iglesiasg: willkommen | 22:07 |
@iglesiasg | lisitsyn, how is your German going? :) | 22:08 |
lisitsyn | iglesiasg: schlecht | 22:08 |
* iglesiasg translates | 22:08 | |
@iglesiasg | hehe why so? | 22:08 |
lisitsyn | iglesiasg: well I am having some time management troubles | 22:13 |
lisitsyn | ;) | 22:13 |
lisitsyn | iglesiasg: I have absolutely no idea how sonne|osx cope with that | 22:13 |
-!- travis-ci [~travis-ci@ec2-23-23-31-73.compute-1.amazonaws.com] has joined #shogun | 22:16 | |
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/11600365 | 22:16 |
-!- travis-ci [~travis-ci@ec2-23-23-31-73.compute-1.amazonaws.com] has left #shogun [] | 22:16 | |
-!- zxtx [~zv@149.165.244.2] has joined #shogun | 22:35 | |
lisitsyn | oh gosh wiking you have done crazy amount of work with cmake | 22:40 |
besser82 | lisitsyn: araound? | 22:40 |
lisitsyn | besser82: yes what's up? | 22:40 |
besser82 | lisitsyn: I've seen you're adding ViennaCL to tapkee... | 22:41 |
lisitsyn | besser82: I've made some experiments | 22:41 |
besser82 | lisitsyn: Any motivation to bring this up to SHOGN? | 22:41 |
lisitsyn | besser82: well I should update tapkee in shogun yeah | 22:41 |
lisitsyn | besser82: the problem is that speedup is not really worth the game | 22:42 |
besser82 | lisitsyn: for tapkee or shogun? | 22:42 |
lisitsyn | besser82: the only thing I have seen is some 1.5x speedup with diffusion maps | 22:42 |
lisitsyn | besser82: for shogun it makes sense | 22:42 |
lisitsyn | but needs proper design | 22:43 |
besser82 | lisitsyn: eigenvectors do get really fast with it :) | 22:43 |
lisitsyn | besser82: dense? | 22:43 |
besser82 | lisitsyn: both sparse and dense | 22:43 |
lisitsyn | besser82: full eigendecomposition? | 22:43 |
besser82 | lisitsyn: not tried yet | 22:43 |
lisitsyn | besser82: I mean is it full or partial? | 22:43 |
besser82 | lisitsyn: should be full | 22:44 |
lisitsyn | besser82: that's the critical point because in tapkee full eigendecomposition is never used | 22:44 |
lisitsyn | say you are computing LLE embedding | 22:44 |
lisitsyn | you have N x N sparse matrix | 22:44 |
lisitsyn | but to compute 2d embedding you need *only* 3 smallest eigenvectors | 22:44 |
shogun-buildbot | build #1510 of cyg1 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/cyg1%20-%20libshogun/builds/1510 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 22:45 |
besser82 | lisitsyn: so partial... | 22:45 |
lisitsyn | yeah so I need partial | 22:45 |
lisitsyn | that's why I am tied to arpack :) | 22:45 |
besser82 | lisitsyn: but ViennaCL should support both | 22:45 |
lisitsyn | let me check | 22:45 |
lisitsyn | I don't mind to put it if it is here | 22:45 |
lisitsyn | I tried to make it modular | 22:45 |
lisitsyn | besser82: true, lanczos is here | 22:46 |
besser82 | lisitsyn: :D | 22:46 |
lisitsyn | besser82: ah one more problem - I don't have any gpu ;) | 22:47 |
besser82 | lisitsyn: pocl is your freind :) | 22:47 |
besser82 | lisitsyn: pocl is openCL on cpu only :) | 22:47 |
lisitsyn | besser82: well no, my cpu supports opencl natively | 22:47 |
lisitsyn | that's sandy bridge | 22:47 |
lisitsyn | but hey that's essentially the same | 22:47 |
lisitsyn | I mean I can't see the speedup | 22:47 |
besser82 | lisitsyn: because of no GPU :) | 22:48 |
besser82 | lisitsyn: GTX285 SLI nearly oupterforms recent XEON !!! | 22:48 |
besser82 | lisitsyn: and is really affordable $ | 22:48 |
lisitsyn | besser82: okay thanks for the tip, I'll try to put lanczos to tapkee | 22:48 |
besser82 | lisitsyn: :D | 22:48 |
lisitsyn | besser82: well I am kind of mobile so I used to use notebook | 22:49 |
lisitsyn | I am switching cities sometimes :D | 22:49 |
lisitsyn | like today | 22:49 |
besser82 | lisitsyn: me, too. But mine has GTX580 SLI :D | 22:49 |
besser82 | lisitsyn: in notebook :D | 22:49 |
lisitsyn | besser82: .. and your notebook weights like 8 kg I guess ;) | 22:50 |
besser82 | lisitsyn: nope ~5 kg :) | 22:50 |
lisitsyn | oh damn | 22:50 |
besser82 | but with 17" screen :D | 22:50 |
besser82 | and 3 HDD | 22:50 |
besser82 | if i need ultra-mobility i use my old q45 from samsung ---> ~1.5 kg | 22:51 |
lisitsyn | well you can save some money on fitness carrying tihs bag of potato | 22:51 |
lisitsyn | :D | 22:51 |
besser82 | sure :D | 22:51 |
lisitsyn | besser82: anyway I just don't need gpu computing | 22:52 |
lisitsyn | that's the main reason | 22:52 |
lisitsyn | :) | 22:52 |
besser82 | if you don't need then all is fine :) | 22:52 |
lisitsyn | I played with it out of curiousity but it doesn't look I need it | 22:52 |
besser82 | but in shogun | 22:52 |
besser82 | it would be a real option :D | 22:52 |
lisitsyn | I don't mind to spend some time on that if there is real demand | 22:53 |
besser82 | combined with sth. like openShift I it would be great... | 22:53 |
-!- zxtx [~zv@149.165.244.2] has quit [Ping timeout: 240 seconds] | 22:53 | |
lisitsyn | besser82: I see a few problems there | 22:53 |
besser82 | lisitsyn: which? | 22:54 |
lisitsyn | 1) I don't know how to design heterogeneous software | 22:54 |
lisitsyn | 2) I don't know how to deal with limited memory here | 22:54 |
lisitsyn | I mean say you have kernel svm | 22:54 |
lisitsyn | you need to have all the features in gpu memory | 22:54 |
lisitsyn | otherwise you'll spend most of the time copying stuff back and forth | 22:55 |
besser82 | won't 1.5 Gbytes be enough? | 22:55 |
lisitsyn | besser82: 1.5gb seem to be enough yeah | 22:55 |
shogun-buildbot | build #18 of debian wheezy - memcheck is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/18 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 22:55 |
besser82 | lisitsyn: any recent gtx285+ has at least 1gig | 22:56 |
lisitsyn | oh I have no clue about these numbers ;) | 22:56 |
lisitsyn | well yeah 1gb should be ok | 22:56 |
besser82 | lisitsyn: gtx285 is from ~ 4years ago ;) | 22:56 |
besser82 | lisitsyn: today's GTX{5,6,7,8}** are all <= 2Bytes of mem | 22:57 |
lisitsyn | besser82: I haven't had any non-notebook computer nearly the same time | 22:57 |
besser82 | lisitsyn: today's GTX{5,6,7,8}** are all >= 2Bytes of mem | 22:57 |
lisitsyn | like once I started my studies | 22:57 |
lisitsyn | heh | 22:57 |
besser82 | desktops have these specs as well | 22:58 |
lisitsyn | besser82: I have to get back to the fact I don't need it :D | 22:58 |
besser82 | lisitsyn: ok, ok :D | 22:58 |
lisitsyn | besser82: but if I get some hardware like that I'll try it for sure | 22:58 |
lisitsyn | besser82: we should employ gpu to compile shogun to make it fast :D | 23:06 |
besser82 | lisitsyn: using gpu on compiling or compiling with support for gpu-use? | 23:07 |
lisitsyn | besser82: jk :) it is so slow | 23:08 |
besser82 | lisitsyn: jk? | 23:08 |
lisitsyn | besser82: just kidding | 23:08 |
besser82 | lisitsyn: kk | 23:08 |
besser82 | lisitsyn: but compiling isn't that slow takes ~7 minutes | 23:09 |
besser82 | lisitsyn: try compiling Mahout :D | 23:09 |
besser82 | lisitsyn: will run ~40 minutes on i7 | 23:09 |
lisitsyn | besser82: really? | 23:12 |
besser82 | lisitsyn: yes | 23:12 |
lisitsyn | besser82: but its java, how can it be that slow? | 23:13 |
lisitsyn | I am totally surprised | 23:14 |
besser82 | lisitsyn: because of maven | 23:14 |
besser82 | lisitsyn: hadoop takes ~1h | 23:14 |
lisitsyn | besser82: what makes it that slow with maven? | 23:14 |
besser82 | lisitsyn: not sure... | 23:15 |
besser82 | lisitsyn: no idea | 23:15 |
lisitsyn | besser82: sounds so weird! | 23:15 |
lisitsyn | I am really surprised because I was used to think java = rather fast compilation | 23:15 |
besser82 | lisitsyn: not even close :D | 23:16 |
lisitsyn | C++ is terribly slow when it comes to expression templates like in eigen3 | 23:16 |
besser82 | lisitsyn: that's true, but java is always slow :) | 23:17 |
lisitsyn | well at my job we have quite good performance with java | 23:18 |
lisitsyn | we are heavily using message passing with akka and I can't say it is slow | 23:18 |
lisitsyn | well and I used to compile terribly big modules at my previous job and it was ok too | 23:19 |
lisitsyn | that's why I am surprised | 23:19 |
besser82 | lisitsyn: I really don't know what the apache-people did to it, making it that slow :( | 23:20 |
lisitsyn | besser82: when I hear apache I think of cemetary | 23:20 |
lisitsyn | like they have zillions of dead projects | 23:21 |
besser82 | lisitsyn: but Mahout is one of the recent ones :) still in beta | 23:21 |
besser82 | lisitsyn: or RC now | 23:21 |
lisitsyn | yeah that's not dead | 23:21 |
lisitsyn | I have used http://commons.apache.org/proper/commons-scxml/ | 23:22 |
lisitsyn | but I didn't know what to think with last commit in ~2007 | 23:22 |
lisitsyn | sonne|osx: did you mean both integration and unit tests? | 23:27 |
-!- besser82 [~besser82@fedora/besser82] has quit [Quit: freedom, friends, features, first ---> fedoraproject.org] | 23:30 | |
shogun-buildbot | build #25 of clang34 - static analysis is complete: Failure [failed analyse] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20static%20analysis/builds/25 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 23:33 |
shogun-buildbot | build #12 of clang34 - undefined behaviour analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20undefined%20behaviour%20analysis/builds/12 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 23:43 |
shogun-buildbot | build #16 of clang34 - thread analysis is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/clang34%20-%20thread%20analysis/builds/16 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 23:57 |
--- Log closed Sat Sep 21 00:00:08 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!