--- Log opened Wed Aug 31 00:00:06 2011 | ||
--- Day changed Wed Aug 31 2011 | ||
shogun-buildbot | build #164 of python_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org:5347/builders/python_modular/builds/164 | 00:00 |
---|---|---|
serialhex | blackburn & sonney2k... i think there was a mistake with some of my programs and accidentally changed a few python examples, but i changed them back so there should be nothing wrong... | 00:13 |
blackburn | serialhex: np, I already restored python examples | 00:13 |
serialhex | oh, ok :) | 00:13 |
serialhex | thanks blackburn | 00:13 |
blackburn | hope we will manage to release today(for me)/tomorrow(for you) | 00:14 |
blackburn | :) | 00:14 |
serialhex | sweet!!! | 00:15 |
blackburn | serialhex: I'm currently postponing all the trac tickets :D | 00:16 |
serialhex | ??? whats that mean? we're just doing bugfixes & patches now?? | 00:17 |
blackburn | serialhex: sure | 00:20 |
blackburn | serialhex: well we do that at least during the last week | 00:20 |
blackburn | or more | 00:20 |
serialhex | ...also, it seems that the modshogun.rb file in the interfaces/ruby_modular dir i've made is not being copied to the ruby load dir, which is causing the tests to fail :( | 00:21 |
blackburn | serialhex: hmm why? | 00:21 |
blackburn | serialhex: what is the ruby dir? | 00:22 |
blackburn | err load dir | 00:22 |
blackburn | serialhex: ah! forgot, what's up with our strings problem? | 00:22 |
serialhex | well, i checked to make sure, but if you look here: http://www.shogun-toolbox.org:5347/builders/ruby_modular/builds/157/steps/compile/logs/stdio | 00:22 |
serialhex | there is no modshogun.rb file. but the modshogun.so file gets copied... | 00:23 |
serialhex | stings problem??? the thing i e-mailed soeren & baozeng last nightabout? | 00:23 |
blackburn | serialhex: ain't it sufficient to have modshogun.rb in the examples dir? | 00:23 |
blackburn | serialhex: yeah that nightly shit | 00:24 |
serialhex | well, the modshogun.rb file i've been building so that it's easier to use shogun with ruby... making it more ruby-like in essence | 00:24 |
serialhex | erm... lemme forward you the e-mail | 00:25 |
serialhex | OOOHHHH!!!!! the last nite thing was solved by soeren because one needs to call f.close() or you get errors... and that is something i can make more ruby-ish (using blocks) | 00:26 |
serialhex | ...or is there something else???? i think it's all been solved... | 00:27 |
blackburn | serialhex: it was about histogram (not enough bits) | 00:28 |
blackburn | is it solved? | 00:28 |
serialhex | OH! what we talked about at like 4am your time :D | 00:28 |
blackburn | yeah | 00:28 |
serialhex | yes it is!!! soeren solved it, and i guess your estimation of the problem was right! | 00:29 |
blackburn | ah I see | 00:30 |
blackburn | fuck yeah! killed a mosquito | 00:31 |
serialhex | AWESOME1!!!! | 00:32 |
serialhex | hmm..... | 00:33 |
blackburn | serialhex: so lets resolve modshogun issue | 00:34 |
blackburn | serialhex: won't it work with modshogun.rb in the examples folder? | 00:34 |
blackburn | it is the way we do it in python/java/etc | 00:34 |
serialhex | yeah, i'm about to look in the configure file & see where the 'make install' command for ruby is and include that file to be copied | 00:35 |
serialhex | it will work... but if people want to use code like those examples then it won't... it'll throw the same errors | 00:36 |
blackburn | serialhex: it is ok I think | 00:36 |
blackburn | serialhex: I would suggest you to put modshogun.rb right in the ruby_modular examples folder | 00:37 |
blackburn | python_modular won't work without some load.py too | 00:37 |
blackburn | and java_modular without Load.java | 00:37 |
serialhex | hmm, well then moving it to the examples dir & changing "require 'modshogun'" to "require './modshogun'" will fix the problem... | 00:37 |
blackburn | serialhex: wait but why ./modshogun.so? | 00:39 |
blackburn | isn't in in some ruby path? | 00:39 |
blackburn | globally available I mean | 00:39 |
serialhex | hmm... maybe i ought to just rename it to 'shogun_helpers.rb or something, and require that... | 00:40 |
serialhex | ' | 00:40 |
blackburn | serialhex: it is even more convenient - if you want a standalone shogun using .rb - you have to find modshogun.rb somewhere | 00:41 |
blackburn | and when it is in examples dir - you could just copy that and nothing more | 00:41 |
serialhex | well, that was the idea... to make a modshogun.rb file with all the ruby-ish additions, and have it in the same place as modshogun.so, and have the first require the second | 00:42 |
blackburn | serialhex: what is the + of this way? | 00:46 |
serialhex | the end-user dosn't have to do anything extra to get shogun to work like it is in the files | 00:49 |
blackburn | serialhex: so is it ~ to store this file in ruby_modular examples dir? | 00:50 |
serialhex | and as i change things a little here & there on the ruby-side of things, i only have to edit one file, and that will automagically update their ruby configuration when i update ti | 00:51 |
serialhex | it's ok to do that, it's not going to break anything | 00:51 |
blackburn | serialhex: so lets do that? ;) | 00:52 |
serialhex | ok | 00:52 |
serialhex | yeah, i just tried that - i'd have to rename the ruby-ish file to something else and require that seperately to get it to work right... | 01:02 |
blackburn | serialhex: will you make a pull request? | 01:03 |
serialhex | ?? what for? | 01:03 |
serialhex | you want me to do that? i will & do the pull req, but it's gonna take a bit (there are a handful of files to edit) | 01:04 |
blackburn | serialhex: I don't know - I could do that :) | 01:05 |
blackburn | use sed luke | 01:05 |
blackburn | :D | 01:05 |
serialhex | sed?? hmm... another *nix util i do not know how to use :'( | 01:06 |
blackburn | serialhex: wait but there is already shogun_helpers.rb | 01:06 |
serialhex | duh! i think that's 9/10'ths of the modshogun.rb file :P | 01:07 |
blackburn | and load.rb | 01:07 |
blackburn | pretty messy | 01:07 |
serialhex | yes, shogun_helpers.rb is *really* messy... :'( | 01:08 |
serialhex | i need to clean it | 01:08 |
blackburn | serialhex: so the plan is | 01:09 |
blackburn | replace shogun_helpers with | 01:09 |
blackburn | modshogun.rb | 01:09 |
blackburn | then remove load.rb, modshogun.rb | 01:09 |
blackburn | and add to every test require of modshogun.so and shogun_helpers.rb | 01:09 |
blackburn | is it ok? | 01:11 |
serialhex | yes | 01:11 |
serialhex | and i'm in the process of doing that right now :) | 01:11 |
serialhex | ok... how would i use sed to edit all these files? | 01:12 |
serialhex | ^ blackburn? | 01:12 |
@sonney2k | why is modshogun.rb needed? | 01:12 |
blackburn | serialhex: e.g sed -i 's/something/anything' file.rb | 01:13 |
blackburn | oh sorry | 01:13 |
blackburn | sed -i 's/something/anything/g' file.rb | 01:13 |
blackburn | sonney2k: surprise to see you :D | 01:13 |
@sonney2k | blackburn, I didn't manage to fix any of the bugs | 01:14 |
blackburn | sonney2k: I have fixed crash in generator.py | 01:14 |
@sonney2k | blackburn, python examples throwing valgrind errors I cannot make any sense of | 01:14 |
blackburn | eh? all ok now | 01:14 |
serialhex | ...so i could do `sed -i "s/require 'modshogun'/require 'modshogun'\nrequire'shogun_helpers'/g *.rb` ?? | 01:14 |
@sonney2k | blackburn, ?! | 01:14 |
blackburn | serialhex: yeah but I guess you should protect ' with \' or so | 01:15 |
serialhex | ok | 01:15 |
blackburn | sonney2k: what is crashing?? | 01:15 |
@sonney2k | blackburn, running python generator.py works for you? | 01:15 |
@sonney2k | I have valgrind errors all over the place + crashes | 01:15 |
blackburn | sonney2k: have you updated? | 01:16 |
@sonney2k | yes | 01:16 |
blackburn | ehm | 01:16 |
blackburn | what is the errors you get? | 01:16 |
blackburn | are* | 01:16 |
@sonney2k | tons | 01:16 |
blackburn | TParameter? | 01:16 |
@sonney2k | I no | 01:16 |
@sonney2k | no | 01:16 |
@sonney2k | classifier_averaged_perceptron_modular.py | 01:17 |
@sonney2k | try to run that w/ valgrind | 01:17 |
blackburn | UnicodeEncodeError: 'ascii' codec can't encode character u'\xf6' in position 323369: ordinal not in range(128) | 01:17 |
blackburn | I have that error sometimes | 01:17 |
blackburn | no idea how to get there it is | 01:17 |
@sonney2k | I don't have that | 01:18 |
blackburn | sonney2k: any idea how to find that unicode char? | 01:18 |
@sonney2k | is that becaause of you pixel dot again | 01:18 |
blackburn | :D | 01:18 |
blackburn | I don't know I have removed that feature in my vim | 01:18 |
blackburn | it was trailing spaces | 01:18 |
@sonney2k | blackburn, search for ö | 01:19 |
blackburn | u'\xf6' | 01:19 |
@sonney2k | yeah, search for 'ö' | 01:19 |
blackburn | is u ' \xf 6' = ? | 01:19 |
@sonney2k | that is not the problem though... | 01:19 |
blackburn | it crashes compilation on my machine | 01:20 |
@sonney2k | blackburn, which $LANG? | 01:20 |
blackburn | python_modular | 01:20 |
blackburn | sonney2k: I have only python errors when running valgrind | 01:20 |
blackburn | with averaged perceptron | 01:20 |
CIA-3 | shogun: Soeren Sonnenburg master * raff7c0d / src/shogun/preprocessor/KernelPCA.h : remove o umlaut - http://git.io/iRPcZg | 01:21 |
blackburn | oh | 01:21 |
blackburn | umlaut | 01:21 |
@sonney2k | why am I having 'real errors' | 01:21 |
blackburn | :D | 01:21 |
blackburn | sonney2k: e.g.? | 01:21 |
@sonney2k | bernhard's last name | 01:21 |
blackburn | yes I see | 01:21 |
@sonney2k | and btw my boss's name is Müller not Muller ;) | 01:22 |
blackburn | sonney2k: I just copied that from mendeley | 01:22 |
blackburn | sonney2k: so what is the example of 'real error'? | 01:22 |
@sonney2k | uninitialized memory reads | 01:23 |
blackburn | at line? | 01:23 |
serialhex | u / ? whats the difference?? :P | 01:23 |
@sonney2k | serialhex, one sounds like 'you' the other more like the sound of a cow ;-) | 01:24 |
blackburn | serialhex: ? sounds different | 01:24 |
blackburn | :D | 01:24 |
serialhex | i know, i was being facetious (though i didn't know what they sounded like, thanks sonney2k!) | 01:24 |
blackburn | sonney2k: soooo? | 01:25 |
@sonney2k | blackburn, I am compiling | 01:25 |
blackburn | sonney2k: and why the f. you are not sleeping? :D | 01:25 |
@sonney2k | because I am stupid | 01:25 |
blackburn | lie and bad excuse | 01:26 |
blackburn | sonney2k: I have no leaks! | 01:27 |
blackburn | sonney2k: generator works bad but with no segfaults | 01:28 |
blackburn | serialhex: works? | 01:29 |
serialhex | through some cajoling, yes :) there are a couple examples with errors though, fixing them then i'll push | 01:29 |
blackburn | serialhex: okay thanks | 01:30 |
@sonney2k | blackburn, I have these errors | 01:30 |
@sonney2k | AveragedPerceptron | 01:30 |
@sonney2k | ==24399== Conditional jump or move depends on uninitialised value(s) | 01:30 |
@sonney2k | ==24399== at 0xE296750: shogun::DynArray<double>::find_element(double) const (DynArray.h:245) | 01:30 |
@sonney2k | ==24399== by 0xEC2FD39: shogun::CSet<double>::contains(double) (Set.h:68) | 01:30 |
@sonney2k | ==24399== by 0xEC2FC1F: shogun::CSet<double>::add(double) (Set.h:47) | 01:30 |
@sonney2k | ==24399== by 0xEC2F4A9: shogun::CLabels::get_num_classes() (Labels.cpp:122) | 01:30 |
@sonney2k | ==24399== by 0xEC2ECFC: shogun::CLabels::CLabels(shogun::SGVector<double>) (Labels.cpp:42) | 01:30 |
@sonney2k | ==24399== by 0xEE7CF7A: shogun::CLinearMachine::apply() (LinearMachine.cpp:58) | 01:30 |
@sonney2k | I tried to find a bug in there for hours | 01:31 |
@sonney2k | no idea | 01:31 |
blackburn | sonney2k: is it related only to the averaged perceptron? | 01:31 |
@sonney2k | the uninited memory read is in find_element | 01:31 |
@sonney2k | but in the element that is to be found! | 01:31 |
@sonney2k | wait | 01:31 |
@sonney2k | I get an idea | 01:31 |
@sonney2k | I now understaand | 01:32 |
@sonney2k | blackburn, why don't you get this error?!?! | 01:33 |
@sonney2k | you should too | 01:33 |
blackburn | sonney2k: why it appears? | 01:33 |
@sonney2k | it appears in the apply function | 01:33 |
blackburn | of linear machine? | 01:33 |
@sonney2k | where a label vector is allocated | 01:33 |
@sonney2k | yes | 01:33 |
@sonney2k | that label vector is assigned an nr of classes computed | 01:34 |
@sonney2k | by using a set | 01:34 |
blackburn | can't see what is wrong | 01:34 |
@sonney2k | (all labels are put into the set and so one knows how many there are == nr elements in set) | 01:34 |
@sonney2k | some label(s) must not have been assigned proper values | 01:35 |
@sonney2k | that is the only way this can happen | 01:35 |
blackburn | sonney2k: may be it is too late or I am too lame or I am too lamb but I did'n understand | 01:35 |
blackburn | but nevermind just fix it | 01:35 |
@sonney2k | I am too tired | 01:36 |
@sonney2k | tomorrow... | 01:36 |
blackburn | sonney2k: so just let me know what to fix | 01:36 |
blackburn | or fix it tomorrow, no difference | 01:37 |
blackburn | sonney2k: I've cleaned shogun 1.0.0 milestone in trac | 01:37 |
@sonney2k | does all of generator run through for you? | 01:37 |
@sonney2k | library_fisher2x3_modular ERROR generating '../../../testsuite/tests/library_fisher2x3_modular1.txt' using 'library_fisher2x3_modular.py' | 01:37 |
@sonney2k | ? | 01:37 |
blackburn | sonney2k: yes I get these errors | 01:38 |
blackburn | but no segfaults | 01:38 |
@sonney2k | I no longer have segfaults either... | 01:39 |
@sonney2k | weird | 01:39 |
blackburn | sonney2k: most of them run smoothly | 01:39 |
blackburn | sonney2k: heiko's examples are wrong with no parameter list | 01:40 |
blackburn | fisher2x3 have wrong array on input | 01:40 |
blackburn | etc | 01:40 |
serialhex | blackburn, sonney2k: how do i get Modshogun::HDF5File to work? what do i need? | 01:44 |
blackburn | serialhex: I didn't :D | 01:44 |
* serialhex rolls his eyes | 01:44 | |
blackburn | serialhex: python-h5py | 01:45 |
blackburn | I guess | 01:45 |
serialhex | hmm, ok | 01:46 |
blackburn | noonon | 01:46 |
blackburn | serialhex: wrong | 01:46 |
blackburn | I don't know :) | 01:47 |
serialhex | ya know.. your useless blackburn!! :P | 01:47 |
blackburn | sure | 01:47 |
blackburn | okay sleeptime | 01:49 |
serialhex | nite | 01:49 |
CIA-3 | shogun: Sergey Lisitsyn master * r9305ba3 / (3 files): Removed output from python modular examples - http://git.io/v6QpOQ | 01:49 |
blackburn | serialhex good night sonney2k gute nacht | 01:50 |
-!- blackburn [~blackburn@188.168.3.23] has quit [Quit: Leaving.] | 01:50 | |
CIA-3 | shogun: Soeren Sonnenburg master * rffb7c54 / (4 files): fix examples - http://git.io/P34tiA | 01:51 |
@sonney2k | serialhex, libhdf5-serial-dev | 01:51 |
serialhex | awesome! thanks sonney2k, will get that in a min | 01:52 |
@sonney2k | nite hex | 01:53 |
-!- Netsplit *.net <-> *.split quits: @sonney2k | 02:16 | |
-!- Netsplit over, joins: sonney2k | 02:18 | |
shogun-buildbot | build #168 of python_modular is complete: Failure [failed test_1] Build details are at http://www.shogun-toolbox.org:5347/builders/python_modular/builds/168 blamelist: sonne@debian.org | 03:03 |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has quit [Read error: Connection reset by peer] | 09:52 | |
CIA-3 | shogun: Soeren Sonnenburg master * rb17aeae / (3 files in 2 dirs): fix remaining examples - http://git.io/KciTbA | 10:07 |
-!- mode/#shogun [+o sonney2k] by ChanServ | 10:21 | |
CIA-3 | shogun: Soeren Sonnenburg master * r3deadd2 / (5 files in 2 dirs): all generated tests work upon generator.py/tester.py now - http://git.io/84XWUw | 10:47 |
-!- blackburn [~blackburn@188.168.4.101] has joined #shogun | 10:47 | |
CIA-3 | shogun: Soeren Sonnenburg master * re3c14a4 / data : require new test data - http://git.io/Ao2U3w | 10:49 |
blackburn | sonney2k: what's up? | 10:51 |
@sonney2k | all tests run now on my machine | 10:51 |
@sonney2k | we still might have valgrind errors though | 10:51 |
@sonney2k | blackburn, err how did you git push the data stuff to github ? | 10:52 |
blackburn | sonney2k: git added it in data dir..? | 10:52 |
@sonney2k | I have committed a number of files in data/ | 10:53 |
blackburn | and? | 10:53 |
@sonney2k | git push says everything up to date !? | 10:54 |
shogun-buildbot | build #206 of libshogun is complete: Failure [failed git] Build details are at http://www.shogun-toolbox.org:5347/builders/libshogun/builds/206 blamelist: sonne@debian.org | 10:54 |
@sonney2k | I mean you don't have commit 0d3c492e6daf9ed8992b6f4a778260163ff50423 in data right? | 10:55 |
@sonney2k | blackburn, why do ruby examples fail? | 10:57 |
blackburn | sonney2k: cause modshogun.rb with helpers isn't being copied during installation | 10:58 |
@sonney2k | can't we get rid of modshogun.rb? | 10:58 |
@sonney2k | what is this file for? | 10:58 |
@sonney2k | btw | 10:58 |
blackburn | sonney2k: it almost the same as load.py | 10:58 |
blackburn | sonney2k: yeah we could rename it to load.rb or so | 10:59 |
blackburn | and when require in every example | 10:59 |
@sonney2k | blackburn, so it can stay in examples as load.rb | 10:59 |
blackburn | yes | 10:59 |
@sonney2k | I mean the loading helper is not really 'nice/fast' code | 10:59 |
@sonney2k | it is only to illustrate how to set matrices etc from $LANG into sg | 11:00 |
blackburn | sonney2k: yeah but what else can we do? | 11:00 |
@sonney2k | keep the load.rb around | 11:00 |
@sonney2k | in examples | 11:00 |
blackburn | yes sure | 11:00 |
blackburn | it is better | 11:00 |
@sonney2k | blackburn, any idea about the data repo? | 11:01 |
blackburn | sonney2k: I have outdated data now | 11:01 |
blackburn | you haven't pushed to shogun-data | 11:02 |
blackburn | sonney2k: https://github.com/shogun-toolbox/shogun-data my commit is the last one | 11:02 |
@sonney2k | blackburn, what is your git remote -v ? | 11:03 |
blackburn | sonney2k: in data? | 11:03 |
@sonney2k | git remote -v | 11:03 |
@sonney2k | githubgit@github.com:shogun-toolbox/shogun-data.git (fetch) | 11:03 |
@sonney2k | githubgit@github.com:shogun-toolbox/shogun-data.git (push) | 11:03 |
@sonney2k | of course | 11:03 |
blackburn | origingit://github.com/shogun-toolbox/shogun-data.git (fetch) | 11:03 |
blackburn | origingit://github.com/shogun-toolbox/shogun-data.git (push) | 11:03 |
blackburn | well the same | 11:03 |
@sonney2k | that's all? | 11:03 |
blackburn | yeah | 11:03 |
@sonney2k | no not the same | 11:03 |
@sonney2k | you cannot push under this url | 11:04 |
blackburn | ah | 11:04 |
blackburn | yes | 11:04 |
blackburn | readonly url | 11:04 |
@sonney2k | blackburn, so how did you ever add data? | 11:04 |
blackburn | sonney2k: let me try now | 11:04 |
blackburn | sonney2k: ohhh I see | 11:06 |
blackburn | sonney2k: checkout to master branch! | 11:07 |
blackburn | you are (as me) on the master branch | 11:08 |
blackburn | errrrrnot on the master branch | 11:08 |
@sonney2k | hmmhh now the new data is gone | 11:10 |
@sonney2k | but you are right | 11:10 |
@sonney2k | will fix the ruby stuff first and then | 11:10 |
@sonney2k | blackburn, did you document all the classes with some @brief stuff? | 11:11 |
blackburn | sonney2k: not yet | 11:11 |
@sonney2k | then please... | 11:11 |
blackburn | okay | 11:11 |
blackburn | but after breakfast :D | 11:11 |
blackburn | brb | 11:12 |
shogun-buildbot | build #170 of python_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org:5347/builders/python_modular/builds/170 | 11:15 |
CIA-3 | shogun: Soeren Sonnenburg master * r52fb28d / (18 files in 3 dirs): fixed a number of .rb scripts and removed broken non-trivial to fix ones - http://git.io/fKEtqw | 11:25 |
shogun-buildbot | build #207 of libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org:5347/builders/libshogun/builds/207 | 11:29 |
blackburn | here | 11:29 |
CIA-3 | shogun: Soeren Sonnenburg master * rb23e6ba / data : require new test data - http://git.io/UiXSbg | 11:31 |
blackburn | sonney2k: could you remind me how to get list of non-@briefed classes? | 11:31 |
@sonney2k | cd examples | 11:31 |
@sonney2k | make | 11:31 |
@sonney2k | blackburn, alright new data should be in | 11:32 |
@sonney2k | python / ruby stuff now works | 11:32 |
@sonney2k | (I fixed what I could and removed the rest) | 11:32 |
blackburn | sonney2k: ehh not examples missing doc, @brief | 11:32 |
@sonney2k | blackburn, doc | 11:32 |
@sonney2k | cd examples | 11:32 |
@sonney2k | make doc | 11:32 |
blackburn | blackburn@blackburn-laptop:~/shogun/shogun/examples$ make doc | 11:33 |
blackburn | make: *** No rule to make target `doc'. Stop. | 11:33 |
blackburn | sonney2k: ^? | 11:33 |
@sonney2k | examples? | 11:33 |
@sonney2k | cd doc | 11:33 |
@sonney2k | make | 11:33 |
@sonney2k | btw what happened to the russian docu? did you cancel it? | 11:34 |
blackburn | sonney2k: yes, I told you?! | 11:34 |
@sonney2k | hmmhh maybe you got disconnect when I replied that it is nice to attract new-comers | 11:34 |
blackburn | sonney2k: I have another idea how to attract | 11:35 |
@sonney2k | anyway better do doc now | 11:36 |
blackburn | sonney2k: corrcoef between 'average english' and 'machine learning interested' = 1.0 ;) | 11:36 |
blackburn | for russian folks I mean | 11:36 |
blackburn | sonney2k: did make doc but where to get un@briefed classes list? | 11:38 |
shogun-buildbot | build #168 of ruby_modular is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org:5347/builders/ruby_modular/builds/168 | 11:55 |
@sonney2k | blackburn, open doc | 12:01 |
blackburn | sonney2k: yeah yeah in progress already | 12:01 |
@sonney2k | please continue in brain on mode :) | 12:01 |
blackburn | sonney2k: done | 12:22 |
blackburn | sonney2k: ?? ??????? | 12:23 |
CIA-3 | shogun: Sergey Lisitsyn master * rceee77a / (43 files in 9 dirs): Doc improvements - http://git.io/BkiXIg | 12:34 |
-!- heiko [~heiko@134.91.53.213] has joined #shogun | 13:16 | |
blackburn | heiko: hi how are you? | 13:20 |
heiko | blackburn, hi! I am fine, a bit tired, packing stuff all night, and you? | 13:21 |
blackburn | I am fine, already packed-unpacked :) | 13:21 |
blackburn | but I was moving ~100km, not that much as you :D | 13:21 |
heiko | ok, but a lot of work also :) | 13:21 |
heiko | We will do it on friday | 13:21 |
heiko | and still much stuff to do in the appartment | 13:22 |
heiko | but I want to do some hacking now, perhaps I can finish the migration today, if there do not appear more errors | 13:22 |
blackburn | ??? | 13:22 |
blackburn | wow | 13:22 |
blackburn | thought we postponed that | 13:23 |
heiko | yes we did | 13:23 |
heiko | but I want to try if it works after this last error | 13:23 |
heiko | We will see | 13:23 |
heiko | probably not, experience tells that :) | 13:23 |
heiko | This feature is like a nest of bees you have to dig into | 13:24 |
blackburn | can I help you somehow? | 13:24 |
heiko | I dont know, I think not really unless you want to dig into all the stuff | 13:24 |
heiko | Its nearly working ... but only nearly | 13:24 |
blackburn | I see | 13:24 |
heiko | Well I will try now for two hours and if it does not work then I will continue moving :) | 13:25 |
-!- serialhex [~quassel@99-101-148-183.lightspeed.wepbfl.sbcglobal.net] has joined #shogun | 13:29 | |
@sonney2k | blackburn, ok | 13:31 |
@sonney2k | blackburn, what else do we need? | 13:31 |
blackburn | sonney2k: I don't know | 13:31 |
@sonney2k | blackburn, from my side examples /tests/doc etc work | 13:32 |
@sonney2k | so only thing now is to fix links on website | 13:32 |
@sonney2k | and make release script | 13:32 |
@sonney2k | blackburn, did you check the remaining python tests/examples? | 13:38 |
serialhex | sonney2k: i'm getting thr ruby examples ready, and doing the libsvm minimal example i get this wile training the svm: | 13:40 |
serialhex | [ERROR] assertion kernel->get_num_vec_lhs()==problem.l failed in file classifier/svm/LibSVM.cpp line 85 | 13:40 |
serialhex | classifier_libsvm_minimal_modular.rb:32:in `train': [ERROR] assertion kernel->get_num_vec_lhs()==problem.l failed in file classifier/svm/LibSVM.cpp line 85 (fatal) | 13:40 |
serialhex | from classifier_libsvm_minimal_modular.rb:32 | 13:40 |
serialhex | want me to gist my code and see whats up? | 13:41 |
@sonney2k | I think the number of examples doesn't match the number of labels | 13:41 |
serialhex | ok... | 13:41 |
@sonney2k | I guess matrix has not the rihgt number of columns | 13:41 |
blackburn | sonney2k: not, tested only distance and kernel | 13:42 |
blackburn | one from both | 13:42 |
@sonney2k | then please continue with more complex stuff | 13:42 |
blackburn | sonney2k: did you rewrite serialized tests? | 13:43 |
@sonney2k | ? | 13:43 |
blackburn | sonney2k: testsuite/tests | 13:43 |
blackburn | sonney2k: are they old now? | 13:44 |
@sonney2k | no | 13:44 |
@sonney2k | it would make sense to get them working again | 13:44 |
@sonney2k | these things are independent of our SGVector/subset whatever transition | 13:44 |
@sonney2k | and should all still work | 13:44 |
@sonney2k | not much work either | 13:44 |
blackburn | sonney2k: so may be I could compare outputs? | 13:44 |
@sonney2k | blackburn, you could get make run-testsuite to work | 13:45 |
@sonney2k | these tests compare outputs already | 13:45 |
blackburn | but I thought we compare to old shogun tests? | 13:45 |
@sonney2k | blackburn, we even have another 'old' testsuite in shogun | 13:45 |
@sonney2k | that one just needs adjustments like classify -> apply renames | 13:46 |
@sonney2k | then it will run for all (old) interfaces | 13:46 |
@sonney2k | would be good to get these to work | 13:46 |
blackburn | okay | 13:46 |
blackburn | but a bit later, gtg | 13:46 |
CIA-3 | shogun: Soeren Sonnenburg master * r49d04a5 / (76 files in 9 dirs): rename r -> r_static etc - http://git.io/l_uu1g | 14:20 |
shogun-buildbot | build #210 of libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org:5347/builders/libshogun/builds/210 blamelist: sonne@debian.org | 14:23 |
-!- f-x [75c0d29e@gateway/web/freenode/ip.117.192.210.158] has joined #shogun | 14:33 | |
@sonney2k | serialhex, I would prefer if you keep the helper functions in load.rb or utils.rb in the examples directory | 14:47 |
serialhex | ok, i can change that later on today sonney2k | 15:09 |
-!- heiko [~heiko@134.91.53.213] has quit [Ping timeout: 258 seconds] | 15:47 | |
-!- f-x [75c0d29e@gateway/web/freenode/ip.117.192.210.158] has quit [Ping timeout: 252 seconds] | 16:19 | |
blackburn | sonney2k: how to fix the testsuite? :/ | 16:30 |
CIA-3 | shogun: Sergey Lisitsyn master * rbe81d77 / src/README : Updated interfaces in readme - http://git.io/BYxVqg | 16:51 |
blackburn | sonney2k: make[1]: *** No rule to make target `.depend', needed by `libshogun.so.10.0'. Stop. | 16:53 |
-!- f-x [75c0d29e@gateway/web/freenode/ip.117.192.210.158] has joined #shogun | 17:14 | |
-!- f-x is now known as Guest34307 | 17:14 | |
-!- Guest34307 is now known as f-x_ | 17:15 | |
@sonney2k | blackburn, cd shogun/src | 17:36 |
@sonney2k | make run-testsuite | 17:36 |
blackburn | sonney2k: yeah a lot of errors | 17:36 |
@sonney2k | to the contrary | 17:36 |
@sonney2k | most stuff works still | 17:37 |
blackburn | sonney2k: currently fixing easysvm | 17:45 |
blackburn | sonney2k: so will we release today? | 17:52 |
CIA-3 | shogun: Sergey Lisitsyn master * r2e4b2b5 / (2 files): Fixed esvm at easysvm - http://git.io/dh7yyA | 17:54 |
@sonney2k | blackburn, sure | 17:57 |
blackburn | sonney2k: any urgent thing to do now? | 17:57 |
@sonney2k | blackburn, fix the old test suite | 17:58 |
blackburn | sonney2k: how? what is wrong with it? | 17:58 |
@sonney2k | blackburn, everything that does not print OK needs to be looked at | 17:58 |
@sonney2k | blackburn, if we can show that the old tests work we know that we didn't break anything | 17:59 |
blackburn | sonney2k: what you mean old tests? | 18:07 |
@sonney2k | testsuite/* | 18:07 |
@sonney2k | when you run test_all.sh there | 18:07 |
@sonney2k | err | 18:10 |
@sonney2k | you also have that make[1]: *** No rule to make target `.depend', needed by `libshogun.so.10.0'. Stop. error? | 18:10 |
@sonney2k | hmmhh | 18:10 |
@sonney2k | what happened?! | 18:10 |
blackburn | sonney2k: I don't | 18:10 |
@sonney2k | you don't have that error? | 18:10 |
@sonney2k | or you do have it? | 18:10 |
blackburn | sonney2k: no error there | 18:10 |
blackburn | only on buildbot | 18:11 |
CIA-3 | shogun: Soeren Sonnenburg master * r8e2b540 / testsuite/matlab_and_octave/test_all.sh : ignore tests - http://git.io/oh8SjA | 18:21 |
CIA-3 | shogun: Soeren Sonnenburg master * r6823fdb / (4 files in 3 dirs): SGVectorize dynprog function - http://git.io/V65vPg | 18:21 |
@sonney2k | there was a bug | 18:21 |
@sonney2k | some ../ thing | 18:22 |
@sonney2k | now it should work again | 18:22 |
blackburn | sonney2k: what should work? | 18:23 |
@sonney2k | the Makefile | 18:23 |
shogun-buildbot | build #214 of libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org:5347/builders/libshogun/builds/214 | 18:24 |
blackburn | sonney2k: my preprocessors can't pass this tests | 18:25 |
blackburn | eigenproblems -> slightly different vectors | 18:26 |
blackburn | I have absolutely no idea how to fix *all* these shitty examples | 18:26 |
blackburn | and what is wrong | 18:26 |
@sonney2k | blackburn, come on try hard | 18:37 |
@sonney2k | when sth doesn't work easily skip it and go on | 18:37 |
blackburn | I have no idea where to start | 18:37 |
blackburn | sonney2k: when I run generator/tester all is OK | 18:40 |
@sonney2k | blackburn, ? take an example - if it doesn't work have a look at the src code | 18:40 |
@sonney2k | don't look at generator/tester | 18:40 |
blackburn | where to look?! | 18:40 |
@sonney2k | as I said | 18:40 |
@sonney2k | testsuite/*/test_all.sh | 18:41 |
-!- f-x_ [75c0d29e@gateway/web/freenode/ip.117.192.210.158] has quit [Ping timeout: 252 seconds] | 18:41 | |
@sonney2k | generator/tester work but this is the 'new' testsuite | 18:41 |
@sonney2k | so we cannot rely on it yet | 18:41 |
blackburn | sonney2k: what it runs? | 18:42 |
@sonney2k | the old testsuite | 18:42 |
blackburn | sonney2k: okay and when I got | 18:43 |
blackburn | ../data/kernel/OligoString_DNA_10_3_60_12.mERROR | 18:43 |
blackburn | what should I do? | 18:43 |
@sonney2k | then debug the testsuite | 18:44 |
blackburn | debug what? | 18:44 |
@sonney2k | the test suite | 18:44 |
CIA-3 | shogun: Soeren Sonnenburg master * rfc4625c / (3 files in 2 dirs): further SGVectorization of dynprog - http://git.io/iUaRYA | 18:46 |
@sonney2k | blackburn, clear now? | 18:46 |
blackburn | sonney2k: very, debug the testsuite through debugging the testsuite | 18:48 |
blackburn | no idea what you want | 18:56 |
@sonney2k | blackburn, which interface are you currently testing? | 19:11 |
blackburn | python modular | 19:12 |
blackburn | and most of them have error | 19:12 |
blackburn | ?! | 19:12 |
blackburn | sonney2k: so that implies we broke so much? | 19:19 |
@sonney2k | yes | 19:19 |
CIA-3 | shogun: Soeren Sonnenburg master * r419a67b / (3 files): fix a few regressions - http://git.io/HkOChQ | 19:20 |
@sonney2k | blackburn, well there is no way other than debug the *.py files in the testsuite/python_modular and check what is different | 19:22 |
@sonney2k | blackburn, I will work on the static interfaces for that now | 19:22 |
@sonney2k | but first I have to get home | 19:22 |
blackburn | huh | 19:23 |
blackburn | okay I'll try to fix something | 19:23 |
@sonney2k | so try hard - good luck and cu in ~1 hr | 19:23 |
blackburn | okay.. | 19:23 |
CIA-3 | shogun: Sergey Lisitsyn master * rb5f27b9 / (2 files in 2 dirs): Transitioned copy feature matrix routine to SGMatrix - http://git.io/elTjEw | 19:53 |
@sonney2k | blackburn, back | 21:13 |
@sonney2k | now lets get the work done | 21:13 |
blackburn | sonney2k: yeah ok | 21:20 |
@sonney2k | blackburn, did you manage to reduce the number of issues at all? | 21:22 |
blackburn | sonney2k: not yet but now I know how to work with it | 21:22 |
blackburn | sonney2k: do the release work, I'll ask you if being stucked | 21:22 |
@sonney2k | blackburn, I am currently fixing broken links on the website | 21:22 |
blackburn | okay | 21:23 |
blackburn | sonney2k: ah one issue was fixed | 21:23 |
@sonney2k | I will write some blurb like thanks to google for gsoc etc | 21:23 |
blackburn | yeah you may do this better than me :D | 21:23 |
CIA-3 | shogun: Soeren Sonnenburg master * ra8f57f0 / (4 files in 4 dirs): exclude tests directory form old-tests - http://git.io/AD2v_g | 21:27 |
shogun-buildbot | build #3 of nightly_default is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org:5347/builders/nightly_default/builds/3 | 21:47 |
blackburn | sonney2k: if it says 'accuracy: ..' | 21:50 |
blackburn | what does it mean? | 21:50 |
@sonney2k | then it says sth like there is a deviation above max tolerance | 21:52 |
blackburn | sonney2k: looks like LogPlusOne is broken hah | 21:54 |
blackburn | no idea how it is possible | 21:54 |
@sonney2k | me neither | 21:54 |
blackburn | sonney2k: how much work you have to done now? | 21:55 |
@sonney2k | ? | 21:55 |
@sonney2k | what? | 21:55 |
blackburn | what are the things you have to done now as release process? | 21:55 |
@sonney2k | to do you mean? | 21:56 |
@sonney2k | the make release script | 21:56 |
@sonney2k | I could immediately do a release if I do it manually | 21:56 |
@sonney2k | but it is tedious | 21:56 |
blackburn | fuck | 21:56 |
blackburn | yes | 21:56 |
blackburn | to do | 21:56 |
CIA-3 | shogun: Sergey Lisitsyn master * r16227f5 / testsuite/python_modular/preprocessor.py : Fixed preprocessor testsuite - http://git.io/iw_VmQ | 22:08 |
CIA-3 | shogun: Sergey Lisitsyn master * r7f13b81 / data : Updated data - http://git.io/__h5ew | 22:08 |
blackburn | sonney2k: I'm pretty slow, will you have time to join? | 22:23 |
@sonney2k | I am very slow too (tired ...) | 22:23 |
@sonney2k | so please continue | 22:23 |
@sonney2k | I just wrote some stuff for the webpage / updated / fixed it | 22:23 |
@sonney2k | now 'all' that is missing is the tagging/branching/release magic | 22:24 |
blackburn | sonney2k: I'm not going to stop, but afraid of not manage to complete this work | 22:24 |
@sonney2k | blackburn, I think this will mean we release at 1 or 2am on the day of world peace, i.e. september 1 | 22:26 |
@sonney2k | blackburn, I think we should release with what we have | 22:26 |
@sonney2k | by then | 22:26 |
blackburn | sonney2k: heiko broke kmeans very strange way | 22:26 |
blackburn | now cluster centers is wrong for sure | 22:27 |
@sonney2k | ok | 22:27 |
@sonney2k | then please fix ;) | 22:27 |
blackburn | sonney2k: I think there are a lot of this kind of errors | 22:27 |
* sonney2k continues the release script | 22:27 | |
blackburn | sonney2k: yeah trying to investigate why he did it this way | 22:27 |
blackburn | autumn | 22:33 |
blackburn | heh | 22:33 |
@sonney2k | the autumn of code | 22:51 |
@sonney2k | or better the autumn of release | 22:51 |
blackburn | sonney2k: still stucked with KMeans | 22:53 |
@sonney2k | blackburn, have a look at git log src/shogun/clustering/KMeans.cpp | 23:00 |
@sonney2k | I remember that heiko transitioned distance machines to have vectors to compare with on the lhs | 23:01 |
blackburn | sonney2k: yes but lhs contains full features | 23:01 |
@sonney2k | these are compared against all rhs ones | 23:01 |
blackburn | no idea why | 23:01 |
@sonney2k | not only mu's ? | 23:01 |
blackburn | no, all 33 feature vectors | 23:01 |
@sonney2k | I mean before training it contains traingingdata | 23:01 |
@sonney2k | but afterwards only means or? | 23:02 |
blackburn | and after too | 23:02 |
@sonney2k | hmmmh | 23:02 |
blackburn | sonney2k: hmm in python example all ok | 23:05 |
@sonney2k | great | 23:06 |
blackburn | sonney2k: but in testsuite it goes wrong | 23:06 |
blackburn | so I guess clustering.py issue | 23:06 |
@sonney2k | you mean python test suite? | 23:06 |
@sonney2k | or python_modular one? | 23:06 |
blackburn | python_modular sure | 23:07 |
blackburn | sonney2k: got it | 23:08 |
@sonney2k | I don't understand a thing | 23:09 |
blackburn | sonney2k: distance is being reinited after training | 23:09 |
@sonney2k | ahh | 23:09 |
@sonney2k | true | 23:09 |
@sonney2k | bug | 23:09 |
@sonney2k | hope we don't do it in regular examples | 23:10 |
blackburn | sonney2k: mus thing was better I would say | 23:10 |
@sonney2k | blackburn, no | 23:10 |
@sonney2k | The way it is now mus are on lhs | 23:10 |
@sonney2k | and that way the trained method can just be applied/serialized | 23:11 |
@sonney2k | that wasn't possible before | 23:11 |
CIA-3 | shogun: Soeren Sonnenburg master * rea275ed / data : update data - http://git.io/Vm4E9A | 23:24 |
CIA-3 | shogun: Soeren Sonnenburg master * r4240e55 / Makefile : improved release script for work with git - http://git.io/mksANw | 23:24 |
@sonney2k | blackburn, did you commit? | 23:34 |
@sonney2k | what things are open? | 23:34 |
blackburn | sonney2k: in a min | 23:34 |
@sonney2k | I can hardly continue now | 23:34 |
@sonney2k | I can write some release email in the meantime | 23:34 |
blackburn | sonney2k: yeah write | 23:35 |
blackburn | sonney2k: update your data | 23:36 |
@sonney2k | how? | 23:36 |
@sonney2k | git submodule update ? | 23:36 |
blackburn | I just do git pull in data dir | 23:36 |
blackburn | don't know if it is possible with submodule | 23:37 |
blackburn | you just commited old data | 23:37 |
blackburn | sonney2k: I have moved preproc to preprocessor | 23:37 |
CIA-3 | shogun: Sergey Lisitsyn master * rf6512c7 / testsuite/python_modular/clustering.py : Fixes for clustering tester - http://git.io/AGK6gA | 23:38 |
CIA-3 | shogun: Sergey Lisitsyn master * r67b6b84 / data : Updated data - http://git.io/lgvrpA | 23:38 |
@sonney2k | I didn't do any further commits in data | 23:38 |
blackburn | (01:24:49 AM) CIA-3: shogun: Soeren Sonnenburg master * rea275ed / data : update data - http://git.io/Vm4E9A | 23:38 |
blackburn | sonney2k: ^? | 23:38 |
@sonney2k | blackburn, https://github.com/shogun-toolbox/shogun/commit/ea275ed63d9b654b8a5c709a1ef01a0955d428a5 | 23:39 |
@sonney2k | that's only saying we need new data | 23:39 |
@sonney2k | I think you should do that too | 23:39 |
@sonney2k | git add data | 23:39 |
@sonney2k | git commit data | 23:39 |
blackburn | sonney2k: f4e493... was new data | 23:39 |
@sonney2k | blackburn, you didn't git commit it | 23:40 |
@sonney2k | so no chance I could recognize | 23:40 |
blackburn | sonney2k: https://github.com/shogun-toolbox/shogun/commit/7f13b81348bdee075f0f5b6c7685541790f1b71c | 23:40 |
blackburn | sonney2k: you just did my commit reversed ;) | 23:40 |
@sonney2k | hmmhh why oh why? | 23:41 |
blackburn | cause if you have old data git says that data is possible to commit | 23:41 |
@sonney2k | blackburn, but a git submodule update should fix it | 23:42 |
@sonney2k | but it doesn't here | 23:42 |
blackburn | sonney2k: SVMSGD and SubgradientSVM fail totally | 23:42 |
@sonney2k | blackburn, that's ok | 23:42 |
blackburn | sonney2k: clustering, distances, distributions are ok | 23:42 |
@sonney2k | we changed default loss for svmsgd | 23:43 |
blackburn | sonney2k: some kernels fail (like Oligo) | 23:43 |
@sonney2k | these shouldn't | 23:43 |
blackburn | sonney2k: prunevarsubmean and logplusone too | 23:43 |
@sonney2k | there was no change | 23:43 |
blackburn | regression is ok | 23:43 |
@sonney2k | blackburn, logplusone?! | 23:44 |
@sonney2k | because of preproc rename? | 23:44 |
blackburn | sonney2k: no idea | 23:44 |
blackburn | sonney2k: have you any 'power' to check Oligo and Poly kernel? | 23:45 |
blackburn | ../data/kernel/OligoString_DNA_10_3_60_12.mERROR | 23:47 |
blackburn | km_test: 1.648674e-02, km_train: 1.430996e-02 <--- accuracy: 1.000000e-09 | 23:47 |
blackburn | ../data/kernel/OligoString_DNA_10_4_60_17.mERROR | 23:47 |
blackburn | km_test: 1.691602e-02, km_train: 6.647181e-04 <--- accuracy: 1.000000e-09 | 23:47 |
blackburn | ../data/kernel/Poly_1_0_10_3_6_01_False_False_CUBE_1_FK.mERROR | 23:47 |
blackburn | km_test: 1.489201e-01, km_train: 1.502606e-01 <--- accuracy: 1.000000e-06 | 23:47 |
blackburn | ../data/kernel/Poly_1_0_10_3_6_01_False_False_CUBE_1_TOP.mERROR | 23:47 |
blackburn | km_test: 1.725301e-01, km_train: 1.051410e-01 <--- accuracy: 1.000000e-06 | 23:47 |
blackburn | I'm trying to check logplusone | 23:47 |
@sonney2k | completely broken | 23:48 |
CIA-3 | shogun: Soeren Sonnenburg master * r99966f1 / Makefile : improve release script again - http://git.io/M_M3zQ | 23:58 |
--- Log closed Thu Sep 01 00:00:44 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!