| --- Log opened Mon Oct 17 00:00:01 2011 | ||
| -!- blackburn [~blackburn@188.168.4.242] has quit [Quit: Leaving.] | 01:33 | |
| -!- blackburn [~blackburn@188.168.4.200] has joined #shogun | 09:57 | |
| -!- blackburn [~blackburn@188.168.4.200] has quit [Quit: Leaving.] | 11:40 | |
| CIA-3 | shogun: Soeren Sonnenburg * r00fd1c4 / (2 files): fix svmlight based modular examples to use try/catch - http://git.io/F1vUXA | 12:30 |
|---|---|---|
| shogun-buildbot | build #344 of libshogun is complete: Failure [failed git] Build details are at http://www.shogun-toolbox.org/buildbot/builders/libshogun/builds/344 blamelist: sonne@debian.org | 12:32 |
| CIA-3 | shogun: Soeren Sonnenburg * r3656dd7 / src/README.developer : add comment about newlines to coding conventions - http://git.io/VjatjA | 13:34 |
| CIA-3 | shogun: Soeren Sonnenburg * r42e3e94 / src/configure : add comment on how to use swig's 2.0.4 native python interface - http://git.io/T-sLWw | 15:14 |
| CIA-3 | shogun: Soeren Sonnenburg * rb68cab2 / src/AUTHORS : mention the most active GSoC authors in AUTHORS - http://git.io/k25zNg | 15:28 |
| CIA-3 | shogun: Soeren Sonnenburg * r72d1bde / src/README.data : mention how to obtain data - http://git.io/16kOBQ | 15:44 |
| -!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting] | 16:12 | |
| -!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 16:12 | |
| -!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Client Quit] | 16:14 | |
| -!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 16:14 | |
| CIA-3 | shogun: Soeren Sonnenburg * r002b7bb / src/.version.sh : remove comments - http://git.io/KZkvOQ | 16:15 |
| shogun-buildbot | build #349 of libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/libshogun/builds/349 | 16:23 |
| shogun-buildbot | build #302 of r_static is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_static/builds/302 blamelist: sonne@debian.org | 16:36 |
| shogun-buildbot | build #300 of octave_static is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_static/builds/300 blamelist: sonne@debian.org | 16:43 |
| CIA-3 | shogun: Soeren Sonnenburg * rcb90e0d / (2 files): uninline code in CrossValidation.h - http://git.io/fnA5hA | 19:03 |
| shogun-buildbot | build #303 of r_static is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_static/builds/303 | 19:07 |
| shogun-buildbot | build #301 of octave_static is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_static/builds/301 | 19:14 |
| shogun-buildbot | build #236 of octave_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_modular/builds/236 | 19:28 |
| shogun-buildbot | build #219 of r_modular is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_modular/builds/219 blamelist: sonne@debian.org | 19:48 |
| -!- blackburn [~blackburn@188.168.4.221] has joined #shogun | 20:54 | |
| @sonney2k | blackburn, hmmhh I got an answer at the swig mailinglist | 21:06 |
| blackburn | sonney2k: about? | 21:06 |
| @sonney2k | reducing wrapper size | 21:06 |
| blackburn | ah | 21:06 |
| @sonney2k | basically 'do you really need to include all headers'? | 21:06 |
| blackburn | and what's the question? | 21:06 |
| @sonney2k | so what do we do now? | 21:09 |
| blackburn | :D | 21:11 |
| blackburn | sorry | 21:11 |
| blackburn | question^Wanswer | 21:11 |
| blackburn | all ? | 21:11 |
| @sonney2k | blackburn, | 21:12 |
| @sonney2k | The only real support for this problem is %import with multiple modules. | 21:12 |
| @sonney2k | All target languages should work with multiple modules. Some might not, | 21:12 |
| @sonney2k | but then consider it a bug in a major area which should be fixed. The | 21:12 |
| @sonney2k | -fvirtual and -fcompact options you refer to offer some reduction, but | 21:12 |
| @sonney2k | nothing really large. You could alternatively wrap less code, do you | 21:12 |
| @sonney2k | really need all of the code wrapped? | 21:12 |
| blackburn | sonney2k: I have no idea what we could 'un'wrap | 21:25 |
| shogun-buildbot | build #38 of nightly_default is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/38 | 21:25 |
| @sonney2k | blackburn, grep template * | wc -l | 21:35 |
| @sonney2k | 257 | 21:35 |
| @sonney2k | so if we support instead of 11 datatypes just 2 or 3 | 21:36 |
| @sonney2k | we get down to ~60 %template's | 21:37 |
| @sonney2k | my suggestion would be we add a --datatypes= switch to configure | 21:37 |
| @sonney2k | and default it to very few types | 21:37 |
| @sonney2k | say float/double/uint8 | 21:38 |
| CIA-3 | shogun: Soeren Sonnenburg * r6f2007f / src/shogun/io/File.h : add a get_file_descriptor() function to File - http://git.io/LZUFLA | 22:06 |
| blackburn | sonney2k: +1 | 22:15 |
| blackburn | sonney2k: you could just make it -regular-datatypes -extended-datatypes or so | 22:18 |
| @sonney2k | *arg* our example data has train and test data | 22:19 |
| @sonney2k | but only train label | 22:19 |
| @sonney2k | no *test* ones | 22:19 |
| @sonney2k | :( | 22:19 |
| CIA-3 | shogun: Soeren Sonnenburg * rd651da1 / examples/undocumented/octave_modular/classifier_svm_serialize_modular.m : | 22:22 |
| CIA-3 | shogun: add example for svm serialization and octave. | 22:22 |
| CIA-3 | shogun: thanks Michael Herrmann for the contribution - http://git.io/b4UcgA | 22:22 |
| blackburn | sonney2k: when do you think we can make it 1.1ed? | 22:23 |
| blackburn | heh | 22:23 |
| @sonney2k | what do you think? | 22:23 |
| blackburn | sonney2k: middle of november? | 22:23 |
| @sonney2k | I mean did you check for bugs? | 22:23 |
| blackburn | sonney2k: what exactly? | 22:24 |
| @sonney2k | whether test suite still works? | 22:24 |
| blackburn | sonney2k: hmm I didn't | 22:24 |
| @sonney2k | in principle we could release now | 22:24 |
| blackburn | no-no | 22:24 |
| @sonney2k | at least examples run again and buildbot is happy | 22:24 |
| @sonney2k | but the question is what do we want to have finished | 22:25 |
| @sonney2k | for example | 22:25 |
| @sonney2k | should 1.1. compile under cygwin ? | 22:25 |
| @sonney2k | osx? | 22:25 |
| blackburn | sonney2k: I have two more algos to finish | 22:25 |
| @sonney2k | do we want to limit resources needed to compile shogun? | 22:25 |
| @sonney2k | etc | 22:25 |
| @sonney2k | then there is the .h -> .cpp inline migration | 22:26 |
| blackburn | sonney2k: so how do you estimate date? | 22:26 |
| @sonney2k | still lots of work todo for that | 22:26 |
| @sonney2k | however we could in principle release any time as long as we don't introduce regressions | 22:26 |
| blackburn | sonney2k: well there are always a lot of work | 22:26 |
| @sonney2k | exactly | 22:26 |
| blackburn | but we have to release sometimes :) | 22:27 |
| blackburn | sonney2k: I propose to do: | 22:28 |
| blackburn | reduce size | 22:28 |
| blackburn | check for regressions | 22:28 |
| blackburn | finish two my algos | 22:28 |
| blackburn | hmm.. that's all? | 22:29 |
| @sonney2k | cygwin / osx | 22:29 |
| blackburn | yes | 22:29 |
| @sonney2k | the only problem I see is that we don't have migrations working | 22:29 |
| blackburn | sonney2k: which migrations/ | 22:29 |
| @sonney2k | so your layout change might introduce non readable objects | 22:30 |
| blackburn | ?? | 22:30 |
| @sonney2k | serialization! | 22:30 |
| blackburn | layout change? | 22:31 |
| @sonney2k | converter! | 22:31 |
| blackburn | ehmm how can it produce non-readable objects? | 22:31 |
| blackburn | I don't think someone ever saved say LLE or LTSA :D | 22:31 |
| @sonney2k | if someone uses a preprocessor from your LLE code | 22:31 |
| @sonney2k | lets hope so | 22:31 |
| blackburn | sonney2k: the only thing available to save is k parameter | 22:32 |
| blackburn | :D | 22:32 |
| blackburn | nothing else | 22:32 |
| blackburn | sonney2k: all algos I've implemented are embedding methods, not transforming like PCA | 22:33 |
| blackburn | there is really nothing to save | 22:33 |
| shogun-buildbot | build #302 of octave_static is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_static/builds/302 blamelist: sonne@debian.org | 22:41 |
| blackburn | sonney2k: soon I will get up to that drtoolbox with 343423432432 methods :) | 22:45 |
| @sonney2k | blackburn, I thought that guy over there had ~30 dr methods or so? | 22:45 |
| @sonney2k | or 50 even | 22:46 |
| blackburn | sonney2k: most of them are useless and he treats Isomap and Landmark Isomap as different algos | 22:46 |
| blackburn | sonney2k: I do not have any iterative algos | 22:47 |
| blackburn | and plan to continue this way | 22:47 |
| @sonney2k | what are examples for iterative algos? | 22:49 |
| blackburn | sonney2k: t-sne, sammon mapping | 22:53 |
| blackburn | manifold sculpting, etc | 22:53 |
| blackburn | sonney2k: I believe only in eigenmagic :D | 22:53 |
| @sonney2k | do you know what SERIALIZABLE_DUMMY is? | 22:53 |
| blackburn | ehhh | 22:54 |
| blackburn | no | 22:54 |
| shogun-buildbot | build #237 of octave_modular is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org/buildbot/builders/octave_modular/builds/237 blamelist: sonne@debian.org | 23:04 |
| shogun-buildbot | build #220 of r_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/r_modular/builds/220 | 23:14 |
| CIA-3 | shogun: Soeren Sonnenburg * r06a4899 / examples/undocumented/octave_modular/classifier_svm_serialize_modular.m : fix typo : -> ; - http://git.io/fszKHg | 23:42 |
| CIA-3 | shogun: Soeren Sonnenburg * r58f986e / (6 files in 2 dirs): | 23:42 |
| CIA-3 | shogun: build shogun's interfaces by default only for a limited set of data types | 23:42 |
| CIA-3 | shogun: namely char,uint8_t,uint16_t,uint64_t,float32_t,float64_t - http://git.io/1G_3ww | 23:42 |
| @sonney2k | blackburn, ok we are down to 375kLoC (from 531kLoC) | 23:43 |
| @sonney2k | when compiling with datatypes=char,uint8_t,uint16_t,uint64_t,float32_t,float64_t | 23:43 |
| @sonney2k | but examples use all types... | 23:47 |
| @sonney2k | so it is only a poor mans solution | 23:48 |
| @sonney2k | but yes better than nothing | 23:48 |
| --- Log closed Tue Oct 18 00:00:06 2011 | ||
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!